v0.5.9 🛡️ — the 2026-09-27 rescan fixes

v0.5.9 🛡️ — the 2026-09-27 rescan fixes

🔍 A four-reviewer rescan of 8e64781 (chain 15 live; its genesis work had never been scanned)
plus cargo audit (0 vulnerabilities). Every fix red-first, each half reverted separately, the
red quoted in its commit. Node and wallet only for chain 15 — no wire break, no consensus-rule
change on chain 15's genesis — so it rolls one node at a time.

🔥 High
🚰 LEDGER-1 A faucet Mint was accepted from any key in the validator register, and a
permissionless bond enters it at once: an allowlisted wallet + 1000 RAND could drain
the faucet every epoch and bond toward a third of the stake. The pool now admits a
faucet Mint only from a genesis validator; genesis staking.faucet_minters makes it a
validity rule at the next cut.
🔁 CN-1 A failed or timed-out sync batch backs its peer off, and the picker ranks misses
before claimed height. Before: two silent peers starved a lagging node for ever.
📚 RS-1 A wallet scanning a pruned node resumes at its floor instead of failing; the public
RPC now ends at the archive (obs1), not a pruning validator.

⚙️ Medium
🪣 CN-2 Sync meters and back-off survive a reconnect; a node-wide Blocks budget; Blocks
serving runs off the consensus loop.
📨 CN-4 Consensus gossip is prechecked (known signer, one signature, size caps) before it is
forwarded, and metered by bytes.
🕰️ CN-3 A NotHeld attestation signs the signer's view and counts only after the lock formed.
🌉 C15-1 Genesis bridge.min_inbound_sequence refuses an old chain's already-minted locks
(next cut).

✅ RS-2 refuted: the public RPC never went through F; callers are metered per address.
↩️ Not in: address-sharing (v0.5.10); CN-5, RS-4, B3, pruning-L1 stay open.

---

Linux x86_64 binaries built on E from the tag (RAND_BUILD_SHA=8abf85e):

28a0130c54fa0b50ebb9085be2616ee3283fe629d4bc2058397fe7eb3a7fcce5  rand-node
14b410946d49f3ac73e7b032d09ec805d5337118143190c55f198f808e40c6d6  rand