Observed (0.9.0-beta, hub v3): an issue created while the hub cache was broken (SharedWriter unavailable) lands in SQLite only. After the cache rebuilds, every write to that issue fails with Issue #N not found in v3 state. integrity hydration correctly detects the 1 sqlite-only issue, but --repair — documented as 'tries to re-emit SQLite-only state back to JSON first' — reports re-hydration FROM the reduced state and leaves the hub at N-1 issues; the FAIL persists and the issue stays unwritable (readable locally, refused on write).
Repro: break/remove the hub cache → crosslink quick ... (succeeds, sqlite-only) → repair the cache → crosslink sync → any issue comment <N> → 'not found in v3 state'; integrity hydration --repair does not publish it.
Ask: make the re-emit leg actually publish sqlite-only issues to the agent's hub ref (or provide an explicit issue publish <N>), so an outage-born issue can rejoin the hub instead of freezing.
Observed (0.9.0-beta, hub v3): an issue created while the hub cache was broken (SharedWriter unavailable) lands in SQLite only. After the cache rebuilds, every write to that issue fails with
Issue #N not found in v3 state.integrity hydrationcorrectly detects the 1 sqlite-only issue, but--repair— documented as 'tries to re-emit SQLite-only state back to JSON first' — reports re-hydration FROM the reduced state and leaves the hub at N-1 issues; the FAIL persists and the issue stays unwritable (readable locally, refused on write).Repro: break/remove the hub cache →
crosslink quick ...(succeeds, sqlite-only) → repair the cache →crosslink sync→ anyissue comment <N>→ 'not found in v3 state';integrity hydration --repairdoes not publish it.Ask: make the re-emit leg actually publish sqlite-only issues to the agent's hub ref (or provide an explicit
issue publish <N>), so an outage-born issue can rejoin the hub instead of freezing.