rdmsm4x-changelog-20260830-2117-ticketing-work-claim-leases

rdmsm4x — work-claim leases, atomic dequeue and plan queues in the fleet ticket engine

When: 2026-08-30 20:51 → 21:20 EDT · Host: rdmsm4x · Session: claude (Opus 5)

One-line summary: the fleet ticket engine now has expiring work-claim leases, an atomic work dequeue, advisory plan queues, cross-host handoff through a single claim authority, two new ticket types (decision, spike), and a fix for a comment-serialisation defect that had fragmented 21 comments into unparseable shards.

Files changed

Path	What
`~/dev/issues/bin/ticket`	+1441 lines: ledger, 11 new subcommands, 2 new types, comment repair, store overrides
`~/dev/issues/bin/ticket-mcp`	+223 lines: 11 new MCP tools, updated type/status enums
`~/dev/issues/tests/test_claim_leases_and_plans.py`	new, 55 checks
`~/dev/issues/index.sqlite`	additive migration: 3 tables, 3 nullable columns; journal_mode delete → WAL
`~/.claude/skills/fleet-ticketing/SKILL.md`	new section 0 (claims), new types, new MCP tools

Commits (`~/dev/issues`, branch main, NOT pushed)

9cec73d restore three fixes a later edit silently reverted
32d4190 work-claim leases, atomic dequeue, advisory plan queues, decision/spike types
5864a86 file four findings from the ticketing-lease work

Backups — how to undo

~/dev/issues/.backups/index.sqlite.bak-20260830-2056
~/dev/issues/.backups/ticket.bak-20260830-2056
~/dev/issues/.backups/ticket-mcp.bak-20260830-2056
~/.claude/skills/fleet-ticketing/SKILL.md backup in this session's scratchpad only

To roll back: git -C ~/dev/issues revert 5864a86 32d4190 9cec73d and restore index.sqlite from the backup above. The schema migration is additive, so an older bin/ticket runs against the migrated index unchanged.

Verification evidence

tests/test_claim_leases_and_plans.py — 55/55 checks pass
test_e2e_adversarial_suite.py 89/89 · test_fts5_queries.py 77/77 · test_stale_index_resolve.py 3/3 · test_concurrent_locks.py pass · test_adversarial_suite.py APPROVE
test_jsonrpc_stdio.py 20/27 — identical before and after; pre-existing ticket-mcp defects, filed as ISSUE-20260830-40
All 107 live ticket files parse and round-trip with zero mismatches
pragma integrity_check ok, 111 rows
Demonstrated live: concurrent claim (one winner), heartbeat, expiry, recorded steal, six-way concurrent next with no double assignment, cross-host claim over real ssh, fail-closed on an unreachable authority

Outstanding owner actions

DEC-20260830-01 — how to reconcile 378 tickets on rdmbair15m5 against 107 here. The claim ledger is fleet-wide; the ticket corpus still diverges.
ISSUE-20260830-39 — a peer edit silently reverted a data-loss fix in bin/ticket. Process problem, not covered by ticket-level leases.
ISSUE-20260830-41 — generated ticket HTML loads mermaid from two CDNs, against the self-contained rule in apps/CLAUDE.md §10.
Other fleet hosts still run the old bin/ticket; they will proxy claims here only after they receive this version.