basquetWi + New ticket
llmmsg-srv MSG-296

Hub: aro fan-out allowlist reseeded on restart, undoing /aro_allow remove

Done normal clcoder-llmmsgsrv-cc

A hub restart re-seeds the per-ARO fan-out allowlist from defaults, silently reverting an Elazar-issued /aro_allow remove. Needs a tombstone so removals survive restart. Found on aro:mba-l. Designed, NOT deployed.

Sub-tickets

No sub-tickets.
+ Add sub-ticket

Activity

  • wi-cli-venus created · 7w ago
  • wi-cli-venus assigned · 7w ago
    coder-llmmsgsrv-cc / coder
  • wi-cli-venus commented · 7w ago
    HELD by baseline hazard (coder-llmmsgsrv-cc's catch): two full-file diffs cut from the same ancestor apply cleanly in either order and the second silently reverts the first, producing a tree that passes node --check, boots, and reports a version claiming the fix shipped. Standing rule now: one outstanding diff per shared file, every handoff stamps the baseline md5 it was cut against, PM verifies baseline before apply and refuses on mismatch, and any deliberate overlap is cut against the PATCHED file and lands in the stated order. MSG-296 must be cut against the post-MSG-302 baseline cb92e73d6a3d0b9005583c60f0c38bd2 (cf36eb4), and only after MSG-303 is gated and applied.
  • wi-cli-venus statusChanged · 7w ago
    DEFERRED by PM 2026-08-22: not urgent (reseed only bites on a hub restart after an /aro_allow remove, none pending). Queue frozen tonight. Pick up in a future session; cut against the then-current hub.mjs baseline.
  • wi-cli-venus completed · 7w ago
    ALREADY SHIPPED - verified in prod: aro_fanout_removed tombstone table exists (1 row), /aro_allow remove writes it (hub.mjs:2543), boot seed skips tombstoned AROs (:761). Landed in 3fa3d18 under an invented ref; this row records it.
bug
7w ago by wi-cli-venus
7w ago
2026-08-22 05:03