hub.mjs carries 13+ design-rationale comments citing a work-item numbering scheme that no longer resolves

Found 2026-09-12 while superseding the always-a-PM invariant for MSG-400. hub.mjs comments cite ids in a pre-project-prefix global bwi numbering: WI 454 (:645, :4119, :4963), BWI #505 (:5140, :5152), BWI #535 (:753, :1141, :2811, :3489), BWI #534 (:799), BWI #548 (:832), BWI #561 (:4967), WI #594 (:601, :2798, :3685, :3693), WI #626 (:3609, :4922, :4943), WI #635 (:5424). NONE of them resolve today: bare numeric refs are refused by the wi CLI, the llmmsg-srv project only reaches MSG-401, and the legacy local store basquetWorkItems.sqlite is a zero-byte file dated 2026-07-19. Not checked: whether the bwi server kept a legacy-id mapping column. WHY THIS IS WORTH AN ITEM AND NOT A SHRUG: these are not decorative. Several are RAILGUARDS - :4963 asserts the always-a-PM invariant, :5140 asserts that resolvePm must stay pure, :3609 asserts PM-seat stickiness across prune. A reader who hits one and cannot resolve the citation assumes the rule was ratified somewhere they failed to look, and defers to it rather than re-examining it. That is precisely what happened with the always-a-PM invariant: it survived unexamined until Elazar named it as the storm engine. An unresolvable citation does not read as missing, it reads as authoritative - the same failure class as a stale rule that expires into over-permission. SCOPE: do NOT mass-rewrite these into new ids - most of the underlying decisions are still in force and re-litigating them is not the ask. For each, either (a) map it to a live ref if the bwi server can resolve the legacy number, or (b) replace the citation with the DECISION plus its date and authority in plain text, which survives any future renumbering. (b) is the default. The three railguards above go first because they are load-bearing; the schema and telemetry comments can follow. Blocked behind eq #110 like the rest of the hub queue. Low priority - this is rot, not a live defect.