v1.1.0

qbit-mining-bootstrap 1.1.0 Release Notes

Release date: 2026-08-17

Highlights

• Hardens the found-block landing path end to end: found blocks are submitted
  to the node before pool accounting, a node-accepted block is never
  terminally abandoned over a stale append epoch or a lost submitblock
  acknowledgement, carry-forward balances are maintained in O(recipients)
  instead of O(ledger history), and landing work runs under landing-class
  deadlines behind a degradable barrier.
• Makes the PRISM ledger writer lease fail closed under contention. Heartbeats
  no longer self-fence during accepted-block persistence, heartbeat staleness
  is measured from verification activity, and a lock-blocked renewal that
  races an expiry claim now fails closed instead of proceeding.
• Scales the PRISM tip-refresh and share-ingest pipeline for production hash
  rates: pay-once-per-generation ledger and serialization costs, epoch-scoped
  refresh fanout, converged latest-tip publication, bounded reconcile
  prefetch, and single-flight job builds that resolve or evict abandoned
  promises.
• Makes payout-artifact builds incremental and debounced, and re-lands
  event-driven artifact reuse behind the PRISM_PAYOUT_ARTIFACT_REUSE
  kill switch after the earlier anchor-scoped attempt was reverted.
• Adds the reward-window leaderboard to the public dashboard API. window=reward
  ranks the canonical live PRISM work window using counted share difficulty
  from the latest 8x permissionless-difficulty window and supports exact
  recipient_id lookup, while the legacy window=3h ranking is preserved.
• Adds an opt-in pool-fee-first coinbase output policy
  (PRISM_COINBASE_OUTPUT_POLICY=pool-fee-first), which requires an enabled
  pool fee, keeps that fee out of CTV fanout, and emits a positive fee output
  at coinbase vout 0.
• Adds per-worker rejects observability to the ckpool lane, native-crash
  diagnosability for the PRISM coordinator, progress-aware health checks, and
  watchdog restart recovery in roughly one second instead of 60s or more.
• Expands CI with Postgres production-scale fixture tests, a native Postgres
  client ledger suite, and a live ckpool rejects observability suite.

Mainnet Compatibility

This release continues to pair with qbit v1.0.0 at
7ebcddb622d6e639041f005a189b048ec2a221fe. The QBIT_GIT_REF and
QBIT_GIT_COMMIT defaults are unchanged, so operators on 1.0.0 keep the same
qbit source pin across this upgrade.

The documented AuxPoW chain IDs are unchanged:

• Mainnet: 47
• Public testnet4: 31430

The qbit mainnet genesis hash for this release line is unchanged:

0000000000004d60aa5d46013991d0a0e2995d89ee98e53068ae196d763e79f2

Upgrade Notes

• Configuration is additive. No .env.example key was removed or renamed
  since 1.0.0. The two new behaviour switches are
  PRISM_COINBASE_OUTPUT_POLICY, which defaults to canonical and preserves
  the 1.0.0 coinbase ordering, and PRISM_PAYOUT_ARTIFACT_REUSE, which
  defaults to 1 and can be set to 0 to fall back to full artifact builds.
• The remaining new keys are PRISM tuning and capacity knobs covering block
  submit and landing timeouts, writer-lease TTL, tip-refresh fanout and worker
  pools, Stratum connection and initial-job limits, coordinator file-descriptor
  limits, payout-artifact cadence, CTV broadcaster batching, public read-model
  caching, and health staleness thresholds. They ship with production-facing
  defaults; review .env.example before overriding them.
• The share-ledger schema bundle in crates/qbit-prism/sql/001_share_ledger.sql
  is applied at coordinator startup. This release adds a
  qbit_payout_carry_forward_current summary table with maintenance triggers.
  A database that already holds active carry history is seeded exactly once
  during that first apply, so expect a one-time backfill proportional to the
  existing carry rows before the coordinator finishes starting.
• Review .env.example and docs/mainnet-deployment.md before upgrading a
  production or mainnet stack.
• Keep production images digest-qualified. scripts/check-env.sh is expected
  to reject mutable image references, missing qbit source provenance, and
  missing final chain-state pins in production mode.
• Run make doctor before starting the stack. For PRISM deployments, run
  make prism-self-check after startup and keep private RPC, Postgres,
  metrics, audit, health, volume, and signing-key surfaces off the public
  internet unless intentionally protected by access controls.

Changes Since v1.0.0

This release rolls up every change merged to main after v1.0.0 (#28 through
#35) together with the 1.x.x hardening line (#36 through #120).

Block landing, submission, and accounting

• Fix the found-block landing livelock with O(recipients) carry-forward
  balances, landing-class deadlines, a degradable barrier, and replay
  hardening (#120).
• Submit found blocks to the node before accounting and keep the submitter
  alive under database saturation (#113).
• Never terminally abandon a node-accepted block over a stale append epoch
  (#119), and fix the earlier stale abandon of accepted blocks (#89).
• Close the accepted-block blind spot behind lost submitblock
  acknowledgements (#93) and carry the post-merge review hardening for that
  fix (#94).
• Skip replay window revalidation once the node offer is out.
• Reconcile accepted-share stats in the background and heartbeat block
  dispositions (#110).
• Avoid reconcile single-flight under the accepted-block payout lock (#96) and
  reduce accepted-block payout gate locking (#50).
• Judge ledger confirmation on settlement balances rather than identity labels
  (#87).

Writer lease, share ledger, and audit

• Stop lease heartbeat self-fencing during accepted-block persistence (#114).
• Fail closed when a lock-blocked lease renewal races an expiry claim (#116).
• Measure lease heartbeat staleness from verification activity (#118).
• Release the writer lease before slow shutdown drains (#47).
• Make share-window reads scale with the window rather than ledger history
  (#115) and isolate share hot-path locking (#83).
• Recover gaps in audit share segments (#88) and fix audit-bundle memory
  amplification (#48).

Tip refresh, job delivery, and Stratum

• Make tip-refresh ledger and serialization costs pay-once-per-generation
  (#99), converge refreshes on latest epochs (#101), and cut hashrate-weighted
  staleness in refresh fanout (#98).
• Scale tip refresh, block finalization, and share ingest for the whale ramp
  (#100) and for same-tip job retention (#31).
• Bound the tip-refresh reconcile prefetch join (#109), resolve or evict
  abandoned single-flight job-build promises (#108), and remove the redundant
  payout-ledger-artifact build from the reconcile path (#95).
• Prevent tip publication starvation (#66) and latest-tip publication livelock
  between concurrent refresh paths (#67); prioritize latest-tip bundle builds
  (#69) and reduce shared bundle latency (#58).
• Delay tip authority until replacement work is ready (#55), supersede
  obsolete refreshes promptly (#42), bound refresh build supersession (#57),
  and preempt obsolete fanout waiters (#43).
• Throttle tip-refresh retries and candidate finalize replays (#71), retry
  failed blockwait refreshes (#41), start the refresh failure budget on first
  failure (#51), and exclude coordination-blocked refreshes from the template
  failure budget (#56).
• Remove coordinator hot-path bottlenecks behind mainnet stale-job rejects
  (#40) and reduce tip-refresh latency under concurrent maintenance (#37).
• Flatten the first-job latency tail for new connections (#92), decouple ready
  bundles from representative clients (#46), reclaim cancelled initial-job
  queue capacity (#68), and bound reconnect storms during job startup (#45).
• Fix build cancellation and refresh races (#64) and the client cleanup lock
  leak (#52); keep Stratum ports open across fast restarts (#63); harden
  Stratum against descriptor exhaustion (#30); bound vardiff idle sweeps
  (#53).

Payouts, artifacts, and CTV fanout

• Make payout-artifact builds incremental and debounced (#112).
• Re-land payout-artifact reuse with event-driven validity and a kill switch
  (#107), after the anchor-scoped reuse line (#102, #103) was reverted (#104).
• Shorten payout delivery gate critical sections (#49) and skip payout
  republish on covered block replays (#59).
• Add an explicit pool-fee-first coinbase output policy (#106).
• Skip no-op CTV fanout status writes and immature-row probes (#60) and
  prevent false CTV broadcaster watchdog exits (#35).

Public dashboard API

• Expose the PRISM reward-window leaderboard (#36) and the pending maturity
  total (#28).
• Label CTV fanout payouts correctly in the public payout read model (#29).
• Smooth public hashrate-series points with a trailing window (#65) and reduce
  public dashboard load on the shared Postgres (#72).

Observability, health, and diagnostics

• Add per-worker rejects observability to the ckpool lane (#97).
• Add native-crash diagnosability to the PRISM coordinator (#91).
• Recover watchdog restarts in roughly one second instead of 60s or more
  (#111) and bound the coordination-blocked template refresh watchdog (#90).
• Add progress-aware health checks (#54) and fix delivery health grace (#82).

Operator Notes

• The historical 0.1.x and 1.0.0 release notes remain in doc/ so public
  release history stays inspectable from the repository.
• docs/prism-payout-artifact-measurement.md documents how to measure
  payout-artifact build cost before tuning the reuse and cadence knobs.
• New Make targets in this release: make test-prism-postgres-scale,
  make test-prism-postgres-native-ledger, and
  make test-ckpool-rejects-observability.