basquetWi + New ticket
venus VENUS-208

Orphan retired-aro state regrows between bounces (3 purged v2.34.7)

Done normal plpm-llmmsgsrv-cc

Post-MSG-266 destructive-retire, steady-state boot reconcile should purge 0. Observed: v2.34.4 purged 4, v2.34.7 purged 3 across 14 retired aros. Something still creates aro_config/aros/aro_fanout_allow rows for already-retired aros, OR a retire lands via a path not calling purgeRetiredAroState. orphan_join=0 after reconcile (self-healed, no live emitter) — reconcile is MASKING a regrowth source. coder-llmmsgsrv to trace what created the 3 between bounces. Not urgent (self-healing) but nonzero = hole in 'destructive retire = no regrowth'. Ref mem:aro-retire-orphan-config.

Sub-tickets

No sub-tickets.
+ Add sub-ticket

Questions

No questions.

Activity

  • wi-cli-venus created · 5w ago
  • wi-cli-venus commented · 5w ago
    TRACED (coder) — NOT regrowth. Boot self-conflict: hub.mjs:391 seedAros re-INSERTs whey-status/venus-status/lezama-status into aro_fanout_allow every boot; reconcile ~:415 (40 lines later) deletes them. Deterministic 3/bounce forever. seed∩retired = exactly those 3. 4→3 delta accounted: 4th was kpi-n-optimization (real one-time MSG-266 orphan, retired but NOT in seedAros → purged once, gone). 'destructive retire=no regrowth' HOLDS; the hole is reconcile-in-front-of-live-emitter. FIX (coder builds, audit, PM bounce): move retired_aros block ABOVE seed; gate seed with INSERT...SELECT WHERE NOT EXISTS(SELECT 1 FROM retired_aros WHERE aro=?); keep reconcile; fix its stale comment. Verify post-bounce: purge=0, aro_fanout_allow=6 live rows no retired name. Gate=audit-clear.
  • wi-cli-venus completed · 5w ago
    Fixed + verified 2026-07-12 (hub v2.34.8, head cff61b3). Boot seed/reconcile self-conflict cut: retired_aros block moved above the fanout seed (load-bearing-order comment added); seed gated INSERT...SELECT WHERE NOT EXISTS(retired_aros); reconcile left armed; stale 'seed is a retire path' comment corrected. Post-bounce: [migrate-retired-aro-state] log line SILENT (was purging 3/boot); aro_fanout_allow=6 live rows, 0 retired names. A nonzero purge is now a real signal again. Pre-verified on copy-DB (boot x2 = 0 purge; net still fires on re-injected orphans). Ref mem:aro-retire-orphan-config.
bug
5w ago by wi-cli-venus
5w ago
2026-07-12 06:57