Repository navigation
Before you upgrade
grpc_admin_api_keyis required. Blockchain refuses to start if it is empty, a placeholder
or under 16 characters, and every RPC butHealthGRPCneeds it. Give every client service the
same key and upgrade Blockchain last, or restart them together. (#1755)- Default RPC credentials are removed. With
rpc_user/rpc_passunset, JSON-RPC rejects every
request and Asset admin routes and dashboard login answer 401. Set both insettings_local.conf
before upgrading;teranode-devusers re-runteranode-dev init. (#1790) - Legacy sync requires a
file://temp_store. The legacy service refuses to start on any
other scheme. Only nodes that movedtemp_storeoff the default must change. Legacy sync now
holds up to 100 GiB of downloaded blocks there: provision the disk. (#1699) - The alert service picks its network by
networkand validatesalert_genesis_keys. On
testnet and stn, clearalert_genesis_keys,alert_protocol_idandalert_topic_nameif they
hold mainnet values. An explicit key list needs 3–10 compressed public keys or startup fails. (#1955) utxostore_updateTxMinedStatusis removed. Nodes that set it tofalsenow always run
mined-tagging and its duplicate-transaction check, with no way to turn them off. Delete the
key; the loader ignores it. (#2019)- Six legacy batcher settings are removed:
legacy_storeBatcherSize,
legacy_storeBatcherConcurrency,legacy_spendBatcherSize,legacy_spendBatcherConcurrency,
legacy_outpointBatcherSizeandlegacy_outpointBatcherConcurrency. Delete them. (#1699) - Four more legacy settings are removed:
legacy_blockFailureBackoffBase,
legacy_blockFailureBackoffMaxDuration,legacy_blockPrefetchBufferBytesand
blockvalidation_legacy_below_checkpoint_fail_closed. Delete them. (#1699) subtreevalidation_processTxMetaUsingCache_BatchSizeis removed and..._Concurrencynow
sizes two service-wide pools instead of each call. Delete the first; raise the second if
subtree validation slows down. (#1957)blockassembly_livenessStallTimeoutmeasures progress, not main-loop passes. Catch-ups,
reorgs and block-validation waits now beat while they progress. If you set it, re-size it from
teranode_blockassembly_liveness_heartbeat_age_seconds. (#1832)- Blockchain caps MTP range requests (
blockchain_maxMedianTimePastHeights, default 10000).
Older validator, subtree and block validation services send whole-chain MTP requests, which it
rejects, so they stop validating. Upgrade Blockchain last or raise the cap above chain height. (#1776) - Conflict recovery adds five
utxo.Storemethods and changes the UTXO WAL. Upgrade block
assembly together with the UTXO store and don't downgrade while conflict-recovery records are
pending; they block startup. Follow conflict recovery.
Before upgrading from v0.16.0, stop block assembly and check the conflict WAL is
empty: a v0.16.0 forward intent left in
it keeps block assembly from starting until it is repaired. The v0.16.0conflict_intents_pending
gauge can't tell you; it only counts at startup. (#2023) - PostgreSQL UTXO store: stop previous-release processes before upgrading. One that starts while
the newpx_transactions_conflictingindex builds blocks alltransactionsreads and writes until
the build ends. Out-of-treeutxo.Storeimplementations addConflictingTxHashes. (#2110) - The Aerospike UTXO store adds a
<set>_conflictwal_startedAtindex, at least 16 MiB per node.
Checkindexes-memory-budgethas room first, or the namespace can stop writes. Without
index-create rights, create it by hand (conflict recovery). (#2151) - Script-number overflow is a consensus failure. Earlier builds accept transactions and blocks
this one rejects: upgrade all your nodes together. These transactions are no longer published to
thekafka_txPolicyRejectedConfigtopic. (#1911) - A block with a non-final coinbase is invalid (
bad-txns-nonfinal), as in SV Node. Earlier
builds accept such blocks: upgrade all your nodes together. Miners must build a final
coinbase. (#2017) - Upgrade block assembly no later than the RPC service. A newer RPC service answers
getminingcandidatewithcoinbase=truewith an error against an older block assembly. The
returned coinbase is now the one block assembly mines. (#2027) - Upgrade the validator after propagation, RPC, legacy and subtree validation. A newer validator
answersERR_TX_UNLOCK_FAILED(42) for an accepted transaction whose unlock timed out, and older
callers treat it as a rejection. Applies where the validator runs as its own service. (#1925) - Upgrade subtree and block validation together. A peer's HTTP 429 is now
ERR_SERVICE_RATE_LIMITED(52), notERR_SERVICE_ERROR(59); an older service gets an unknown
code over gRPC. Log alerts onSERVICE_ERROR (59)for 429s must match the new code. (#1802) - Subtree validation produces to the
kafka_txPolicyRejectedConfigtopic. Grant it write
access where Kafka ACLs are on, and upgrade every subtree validator together: an older one
mistakes the new messages for validator rejections. (#2038) - Block assembly needs produce access to the
kafka_txmetaConfigtopic. When that is set,
block assembly connects at startup and exits if the broker stays unreachable. (#1938) - Unmined transactions are refused while the node is
CATCHINGBLOCKS, including a short
catch-up fromRUNNING. Propagation answers 503 andsendrawtransaction-10. Clients
must resubmit once the node isRUNNING. (#2094) - JSON-RPC error codes change. Missing blocks answer
-5, bad heights and arguments-8,
missing inputs-25and not-ready-28(retry), most of them-32603before. Re-check clients
that act on error codes. (#1637, #1953) - Some script failures move between
TX_POLICYandTX_INVALID. A policy-limit failure is
re-run under consensus limits:TX_POLICYif that passes, elseTX_INVALID. Propagation clients
that retry or drop by these codes should re-check; RPC codes don't change. (#2068) - Six heavy Asset routes are limited to
asset_httpHeavyRateLimit(10 req/s per IP): the
/merkle_proof/*routes,/utxos/:hash/json,/subtree/:hash/txs/jsonand/blockgraphdata.
SPV clients and explorers above that get 429: raise the limit or allowlist peers. (#1789) /lastblocksreturns at most 1000 blocks (lower withasset_maxLastNBlocks) and a negative
nreturns 400. Common-ancestor header routes return 400 above 256 locator hashes. Explorers that
page with a largernmust page in 1000s. (#1792)- Peers must not use a larger
subtreevalidation_missingTransactionsBatchSizethan this node.
Larger batches get 413 (asset_maxBatchRecords, default 16384, and the allowlisted signed-body
cap), which fails the peer's catch-up. Align batch sizes or raise the cap. (#1788, #1797) - Asset no longer sends credentials to cross-origin callers by default. Credentialed browser
callers on other origins, the Vite dev dashboard included, must be listed in
asset_corsAllowOrigins. An invalid entry (*,null, a path) stops Asset at startup. (#1793) asset_trustedProxyCIDRs, when set, no longer trusts private proxies implicitly. List
internal ingress ranges, or every client shares the proxy's rate-limit and ban key. (#1793)asset_concurrency_get_legacy_block_readercaps whole block streams; extra requests fail
after 30 s. Re-size or remove it (default now 0, unlimited). Legacy block serving no longer follows
redirects: pointasset_httpAddressat its final address, or SV peers get no blocks. (#1794)- Four Aerospike pruner metrics are removed:
utxo_pruner_parents_skipped_pruned_total,
utxo_pruner_pruned_set_size,utxo_pruner_pruned_set_saturatedand
utxo_pruner_pruned_set_rotations. Remove their queries. (#1969) - Legacy netsync metrics are removed or renamed.
teranode_legacy_peer_server_OnBlockbecomes
teranode_legacy_peer_server_OnBlockOnDisk, the per-phase histograms are gone and
teranode_legacy_netsync_handle_block_directtimes the block commit. Rebuild those panels. (#1699) teranode_blockvalidation_priority_queue_processed_total{priority}now holds
chain_extending,near_forkordeep_forkinstead ofunknown. Update queries that filter
onunknown. (#1997)teranode_asset_http_peer_auth_result_total{result}changes meaning. With an empty
allowlist every well-formed signed request isnot_allowlisted, so alerts onreplay,
bad_digestorbad_siggo quiet: populateasset_peerAuthAllowlistor retarget them. (#1788)- Asset catch-up route 429s move to
scope="heavy_catchup"on
teranode_asset_http_rate_limited_total. Usescope=~"heavy|heavy_catchup". (#1801) - Global-limit 429s leave
teranode_asset_http_request_duration_secondsand
..._response_size_bytes, and theirmethodlabel folds non-standard methods intoother.
Count those 429s withteranode_asset_http_rate_limited_total{scope="global"}. (#1793) - Out-of-tree
utxo.Storeimplementations must addSpendAndCreateMultiand
ParentOutputsForValidation.utxo.DefaultSpendAndCreateMultiprovides the first. (#1875) - Block assembly can fail to start while a skipped-fork mark is unfinished. A restart finishes
it before loading the pool, and fails if the subtree or UTXO store doesn't answer or
BlockAssemblerSkippedForkMarkis malformed: restart once they answer, or clear that key. (#2065) teranode_blockassembly_processing_stuck_totalgains three reasons:
skipped_fork_mark_unfinished,skipped_fork_mark_record,skipped_fork_subtree_lost. An alert
on the total without areasonfilter now fires on them too. (#2065)- Asset block-list endpoints honour
fromHeightwithincludeOrphans. They now page down
instead of returning the chain's top; clients passing both get the blocks at or below
fromHeight. (#2065)
Upgrading from an earlier version than v0.16.0? Read "Before you upgrade" in every release between yours and this one: https://github.com/bsv-blockchain/teranode/releases. v0.16.0 and earlier list their breaking changes in the changelog, which has since been removed from the repo.
What's Changed
- fix(blockchain): require authentication for service RPCs by @icellan in #1755
- MvP-4841 Closes the chain from a redirected peer POST to an unauthenticated blob write by @ctnguyen in #1800
- chore(logging): demote per-item tracing and expected-condition logs to DEBUG by @oskarszoon in #1692
- test(utils): pass TeranodeTestEnv by pointer so go vet stops flagging lock copies by @codenlighten in #1861
- ci(claude-review): switch to Opus 5.5 and bump claude-code-action to v1.0.238 by @oskarszoon in #1880
- fix(utxo/sql): report the TxInpoints error from the batched read path by @codenlighten in #1863
- ci(claude-review): clean up stuck sticky comments and log permission denial shapes by @oskarszoon in #1882
- fix(usql): retry SQLite busy and locked errors reported with extended codes by @codenlighten in #1858
- fix(teranode-cli): reject positional arguments a command does not take by @codenlighten in #1865
- fix(util): quote Postgres connection values so a URL without a password keeps its database by @codenlighten in #1859
- perf(blockassembly): answer MoveForwardBlock before draining the queue, and drain in parallel by @icellan in #1881
- perf(blockassembly): clear the DiskTxMap index shards in parallel by @icellan in #1878
- fix(blockchain): do not cache a negative GetBlockIsMined answer by @ordishs in #1676
- feat(settings): add Asset admission-budget settings keys (no-op) by @icellan in #1787
- fix(asset): stop shipping admin credentials and close the auth fail-open by @icellan in #1790
- fix(legacy): announce new txs to legacy peers parents first by @oskarszoon in #1838
- ci(claude-review): precompute the PR change set against its merge base by @oskarszoon in #1885
- fix(asset): tighten the HTTP middleware chain and state the signature scope by @icellan in #1793
- fix(asset): give the heavy rate limiter a burst that fits peer catchup by @icellan in #1801
- fix(subtreevalidation): make a rate-limited peer recoverable instead of terminal by @icellan in #1802
- test(soak): add a soak test asserting heap and goroutine counts stay flat under sustained load by @sugh01 in #1871
- fix(e2e): scope legacy sync assertion, drop dead bootstrap DNS lookup by @icellan in #1754
- fix(aerospike): read the store batcher timeout from utxostore_storeBatcherDurationMillis by @icellan in #1903
- feat(utxo): hand the UTXO store a block's transactions as one list, from catch-up and legacy sync by @freemans13 in #1875
- ci(claude-review): install bubblewrap for the subprocess env scrub by @oskarszoon in #1900
- perf(blockassembly): bound the leftover-pass lookups so ingest keeps running during moveForwardBlock by @icellan in #1890
- feat(blockassembly): keep the liveness heartbeat beating through long catch-ups and reorgs by @freemans13 in #1832
- fix(asset): stop serving subtrees that have not been validated by @freemans13 in #1745
- fix(kafka): stop the in-memory broker keeping every message it is given by @freemans13 in #1906
- fix(asset): give the shared main-chain fill its own context, and memoize per-block merkle-proof roots by @icellan in #1795
- perf(model): size the block-validation maps to what the block needs by @icellan in #1902
- fix(asset): admit expensive HTTP routes to the heavy rate limiter by @icellan in #1789
- fix(tracing): detach DecoupleTracingSpan from cancellation when tracing is off by @icellan in #1895
- perf(blockassembly): release the DiskTxMap index on Clear by @icellan in #1904
- fix(asset): make the peer-auth replay claim atomic and canonical by @icellan in #1788
- test(blockassembly): give the remainder fan-out goroutine bound real slack by @icellan in #1905
- ci(lint): compile the longtest test files in the lint typecheck by @freemans13 in #1908
- fix(seeder): read the snapshot footer from the stream so a piped snapshot seeds by @freemans13 in #1894
- fix(utxo/sql): retry SQLite's extended BUSY and LOCKED codes in the create path's lock check by @freemans13 in #1907
- fix(utxo): roll back partial spends so parents never name a record-less spender by @icellan in #1376
- ci(fuzz): fix concurrency, timeout, cache key and Go version in the fuzz workflow by @sugh01 in #1860
- fix(blockvalidation): retry rate-limited catch-up subtree and batch fetches by @oskarszoon in #1899
- fix(blockvalidation): don't charge the announcing peer for a 429 on the block fetch by @oskarszoon in #1932
- fix(asset): unicast the cached node status, cap live websockets and bound the p2p relay by @icellan in #1791
- fix(asset): hold streaming permits for the stream, seek past page offsets, and gate reconstructed transactions by @icellan in #1794
- fix(asset): bound the batch routes and stop sizing allocations from request counts by @icellan in #1797
- fix(asset): bound attacker-controlled n and locator-hash parameters by @icellan in #1792
- perf(asset): bucket the block graph in the store by @icellan in #1798
- fix(validator): accept transactions whose two-phase-commit unlock fails by @ordishs in #1925
- perf(model): fingerprint index for block validation's txMap by @icellan in #1926
- fix(docker): repair healthcheck shell syntax and enable zombie reaping by @mboyd1 in #869
- fix(blockassembly): visit every tx when collecting winning and losing txs in reorgBlocks by @icellan in #1927
- build(deps): bump go-subtree to v1.6.1 by @icellan in #1935
- perf(blockassembly): pointer-free in-memory currentTxMap by @icellan in #1930
- fix(utxo): skip winning transactions the store no longer holds instead of exiting by @icellan in #1934
- fix(blockassembly): reset instead of reorgBlocks when reorg depth reaches UTXO retention by @icellan in #1937
- perf(blockvalidation): find duplicate block inputs by sorting instead of a hash set by @icellan in #1928
- fix: honour the stop return of Iter callbacks across split maps by @icellan in #1936
- fix(blockassembly): reject subtree trailers whose conflicting count exceeds the leaf count by @icellan in #1940
- perf(blockassembly): exact sorted hash set for the block tx map by @icellan in #1929
- chore(deps): build GoBDK against bitcoin-sv 1.2.3 by @oskarszoon in #1911
- fix(blockpersister): reset the decode arena per tx and validate existing subtreeData by streaming by @oskarszoon in #1915
- fix(rpc): return the right error code for a missing block, stop leaking the storage layer by @liam in #1637
- test(blockassembly): run the heavy SortedHashSet cases without -race in the PR test job by @icellan in #1956
- feat(legacy,blockvalidation): download legacy blocks from a header cache and a download ledger, stream them to disk, and apply below-checkpoint blocks in one wave by @freemans13 in #1699
- fix(txmetacache,kafka): stop the cache write-back and consumer fan-out starving txmeta ingestion by @icellan in #1952
- fix(rpc,asset,errors): follow-ups from the #1637 review by @oskarszoon in #1953
- fix(blockpersister): rebuild a block file left without its utxo files, and sleep before retrying a failed block by @oskarszoon in #1949
- fix(blockpersister): bound SubtreeDataWriter.pending to the batch errgroup limit by @oskarszoon in #1959
- fix(subtreevalidation): stop a subtree read flood starving Kafka txmeta ingestion by @icellan in #1957
- fix(docs,settings): correct stale k8s doc headers, implement TestNoMissingTags by @icellan in #1568
- fix(subtreevalidation): stop leaking the blockchain-subscribe retry goroutine across tests by @icellan in #1753
- fix(subtreevalidation): retry store misses under fail fast instead of dropping the subtree by @icellan in #1964
- perf(txmetacache): opt-in lock-free txmeta cache with zero-allocation batch reads by @icellan in #1960
- fix(asset): add switches to withhold backend detail from anonymous HTTP callers by @icellan in #1796
- fix(pruner): keep store URL passwords out of logs and the config dump by @freemans13 in #1747
- fix(blockchain): bound gRPC read RPCs, fix MTP width-dependence, add panic recovery by @icellan in #1776
- chore(settings): clean up dead keys and undefined variables in settings.conf by @icellan in #1744
- fix(blockchain): mark the base_URL length limit as urlsafe by @oskarszoon in #1968
- fix(blob/http): keep a blob-store fault from reading as a miss over HTTP by @oskarszoon in #1963
- docs(blockvalidation): state the full window in which the sweep sets subtrees_set by @freemans13 in #1844
- fix(blockchain): implement the two RPCs that were silently returning Unimplemented by @icellan in #1848
- fix(test): retry daemon start when the p2p port is taken between allocation and bind by @icellan in #1782
- fix(utxo): let callers see through the logging wrapper so logging does not switch pruning off by @freemans13 in #1945
- fix(blockchain): call GetBlockHeadersFromOldest by its current name by @icellan in #1977
- perf(utxo/aerospike): cut the pruner's per-parent CPU and allocations by @freemans13 in #1970
- fix(p2p): report the height a peer advertised, not the sync cap by @ordishs in #1605
- docs: add a CHANGELOG recording the audit batch's breaking changes by @icellan in #1891
- perf(subtreevalidation): cut the subtree validator's probe cost and post-block memory spikes by @icellan in #1967
- Generalise settings-doc drift guard to every settings doc; delete dead kafkaWorkers settings by @icellan in #1639
- chore(settings): drop the legacy outpoint batcher keys again by @icellan in #1971
- fix(utxo): align SQL setMined DAH with Lua and backdate unmined_since on unsetMined by @icellan in #1947
- fix(pruner): skip DAH deletion in a cycle where parent preservation failed by @icellan in #1946
- fix(test): fix the settings and pruner tests broken on main by @icellan in #1981
- fix(seeder): detect truncated utxo-headers files in the header read loops by @icellan in #1604
- perf(blockassembly): build a peer block's transaction map while the block is validated by @icellan in #1973
- perf(model): decode subtree meta into shared slabs straight from the read buffer by @icellan in #1979
- fix(blockassembly): build mining candidates on the chain tip during a reorg by @ordishs in #1983
- perf(model): reuse one inpoints buffer per subtree for the duplicate input check by @icellan in #1989
- fix(aerospike): filter-expression DAH matches Lua setDeleteAtHeight, and child DAH follow-ups on every path by @icellan in #1950
- perf(model): skip own-tx lookups in validOrderAndBlessed, one-cache-line node index buckets by @icellan in #1972
- fix(blockvalidation): bound the free 429 on an announced-block fetch by @icellan in #1942
- test: fix two CI flakes, TestHandleWebSocket and the deferred-drain snapshot test by @icellan in #1985
- test(svnode): run the #942 SV node on free ports, retrying a lost bind by @icellan in #1993
- perf(blockvalidation): apply a quick-validation batch as one list through SpendAndCreateMulti by @freemans13 in #1909
- perf(model): batch node index inserts and parent lookups so their cache misses overlap by @icellan in #1994
- fix(blob/s3): detect a missing object by error code, not by message text by @icellan in #1879
- fix(utxo/sql): stop SetConflicting self-deadlocking on SQLite by @icellan in #1497
- Bump gobdk to v1.3.1 (share sighash precompute per transaction) by @oskarszoon in #2001
- fix(subtreevalidation): CheckBlockSubtrees was stripping its error codes at the gRPC boundary, so block validation blamed the wrong party by @freemans13 in #1650
- refactor(utxo/aerospike/pruner): remove the disabled already-pruned parent skip by @freemans13 in #1969
- docs(sync): label sync durations as estimates and add the tail sync by @ordishs in #1999
- docs(miners): scope backup/restore to operators, add K8s rollback by @ordishs in #1998
- fix(ci): SonarQube refused each PR whose target branch moved after the PR was opened by @freemans13 in #2014
- fix(blockassembly): merge queued reset requests before a reset runs, so a validating request is never answered by a plain reset by @freemans13 in #1995
- fix(blockvalidation): retry a block announcement that hit a local fault instead of losing it by @freemans13 in #2010
- fix(alert): select alert genesis keys and protocol ID by network by @galt-tr in #1955
- fix(utxostore): remove the setting that could switch off mined-tagging and its duplicate-transaction check by @freemans13 in #2019
- test(multinode): retention/limits knobs and blaster fixes for the compose rig by @sugh01 in #2024
- chore: ignore local .mcp.json MCP server config by @ordishs in #1680
- fix(blockvalidation): refuse a legacy block with a missing parent instead of reporting success by @freemans13 in #2021
- test(utxo): cover the external-blob fallback corners, correct two stale doc comments by @icellan in #1996
- fix(blockvalidation): a failed revert of an invalid block's mined tags no longer marks the block done by @freemans13 in #2016
- feat(ui): colour the Network tab chain tip hashes by @ordishs in #2026
- fix(blockvalidation): a panic in a priority worker, the catchup consumer or the Kafka block handler no longer kills the node by @freemans13 in #2012
- Bump gobdk to v1.4.0 (hardware SHA-256, cancellable validation) by @oskarszoon in #2030
- docs(testing): map test targets to levels and state the CI cadence by @ordishs in #1986
- fix(subtreevalidation): charge a peer for subtree data that does not parse, and pin the go-bt reject of a non-minimal CompactSize prefix by @freemans13 in #1489
- fix(subtreevalidation): check a block's conflicting transactions in one batch, not one walk each by @oskarszoon in #2040
- fix(blockvalidation): never charge a peer for a local processing error by @icellan in #2022
- fix(subtreevalidation): keep a peer subtree's conflicting txs aside until a block references it by @icellan in #2038
- fix(blockchain): don't cache a not-found GetBlockExists result by @icellan in #2054
- fix(blockassembly): filter the queue backlog against every block of a multi-block catch-up by @icellan in #2056
- fix(blockassembly): resize queued subtrees in place when they exceed blockmaxsize by @icellan in #1889
- fix(subtreevalidation): source the cached-subtree conflict check from the store by @icellan in #1938
- test(netsync): retry the realTrunk seed when it loses the SQLite table lock by @icellan in #2061
- fix(subtreevalidation): a proven conflict rejects the block even when the check failed elsewhere by @oskarszoon in #2052
- test(settings): keep a local settings_local.conf out of the mmap and txMap default tests by @icellan in #2047
- test(daemon): leave setting blocks mined to the daemon's BlockValidation service by @icellan in #2059
- fix(blockvalidation): don't re-download blocks catchup finds already stored by @icellan in #2053
- fix(blockvalidation): keep mining while a slow block validates, retry instead of dropping by @icellan in #1997
- fix(subtreevalidation): validate peer subtrees skipped during catchup once RUNNING again by @icellan in #2055
- fix(validator): classify policy-mode script errors by a consensus re-run by @icellan in #2068
- docs: describe 1m tps as a design target, not a guarantee by @ordishs in #2032
- fix(blockvalidation): a panic in an errgroup goroutine no longer kills the node by @freemans13 in #2035
- fix(blockassembly): keep a mining candidate's coinbase on its job, so a solution without one can be mined by @freemans13 in #2027
- fix(blockvalidation): only keep subtrees_set on a quick-committed block whose subtree files exist by @freemans13 in #2011
- fix(blockvalidation): reject blocks whose coinbase is not final, and stop block assembly mining them by @freemans13 in #2017
- docs(agents): one test file per production file, and search production code first by @freemans13 in #2076
- fix(blockvalidation): catchup aborts when the first offered header does not link to the common ancestor by @freemans13 in #2034
- fix(blockvalidation): invalidate and publish on an INVALID ancestor verdict during revalidation by @freemans13 in #2033
- fix(blockvalidation): keep a short catch-up's fork transactions in the pool, and stop fetching less-work chains by @freemans13 in #2075
- fix(subtreevalidation): follow-ups to the conflict-check verdict: cancellation wrap, over-budget walk, ctx guard by @oskarszoon in #2066
- fix(utxo): stop a failed ProcessConflicting rollback wedging block assembly by @ordishs in #2023
- fix(blockchain): repair on_main_chain in-process after a failed flag rebuild by @freemans13 in #2049
- fix(blockvalidation): check difficulty of every catch-up header before fetching blocks by @icellan in #2084
- fix(blockassembly): replay every change of the best tip so a branch that briefly won is never skipped by @freemans13 in #2074
- fix(legacy): run the ban listener while the server is up, so bans disconnect peers and stop leaking goroutines by @oskarszoon in #2098
- fix(legacy,propagation,validator,rpc): refuse unmined transactions while the node syncs by @freemans13 in #2094
- fix(legacy): bound the goroutines spent on transaction inventory and relayed transactions by @freemans13 in #2045
- fix(legacy): drop the per-peer requestedTxns delete so main compiles again by @freemans13 in #2106
- perf(blockassembly): sort the unmined reload with a parallel radix sort by @icellan in #2105
- test(model): pin that CVE-2012-2459 mutant and honest bodies share a hash and only the mutant fails Block.Valid by @freemans13 in #1504
- fix(blockassembly): wait for pending mined marking before restamping a stale branch by @freemans13 in #2107
- feat(utxo): add utxoset, a delete-on-spend UTXO store by @freemans13 in #1663
- build: turn off Multipath TCP on listeners by default (CVE-2026-98432) by @oskarszoon in #2112
- fix(utxo): clear parent refs before deleting a tx, and keep reorg-deleted children for recovery by @ordishs in #2102
- fix(legacy): bump go-wire to v1.5.0 and bound peer message reads by @icellan in #2086
- fix(settings): forward alternativeContext for three duration settings by @codenlighten in #1847
- fix(security): keep peer HTTP error bodies and IPs out of public peer state by @icellan in #2109
- fix(subtreevalidation): read conflicting transactions from an index, not every tx of an already-validated subtree by @icellan in #2110
- chore(sonar): exclude .github/scripts from coverage by @oskarszoon in #2125
- docs(changelog): list only changes an operator must act on by @oskarszoon in #2116
- chore(ci): label and close PRs with no human activity, disabled by default by @oskarszoon in #2117
- fix(p2p): pin go-libp2p fork that releases scope of conns closed before Accept by @oskarszoon in #2122
- fix(utxo): floor deletion dates at the mined height and reconcile Aerospike child pages by @ordishs in #2009
- fix(blockassembly): stop re-marking the whole mempool on every reorg by @icellan in #2065
- fix(daemon): exit non-zero when a service fails by @oskarszoon in #2149
- fix(utxo): find pending conflict intents without a namespace scan at startup by @oskarszoon in #2151
- fix(subtreevalidation): leave GOMEMLIMIT room for off-heap memory and decode own external blobs exactly by @oskarszoon in #2177
Full Changelog: v0.16.0...v0.16.1