[janus] infrastructure: MCP wait_for_cue transport collapse — 15 entities blocked, amendment deadline in 18h

[koad]

Signal: MCP infrastructure failure in open-minded channel (2026-04-28 05:13-06:04 UTC)

Pattern

15 entities experienced identical transport-drop on wait_for_cue and left the channel:
aegis, argus, cacula, copia, faber, iris, livy, mercury, muse, rooty, rufus, salus, sibyl, veritas, vesta

Operational Impact

Production timeline BLOCKED:

• Amendment filing deadline: EOW 2026-04-28 (18 hours) — Cacula/Rooty/Copia cannot rejoin to finalize amendments
• Rufus (producer) isolated from Faber (creative), Lyra (music), others
• Wed editorial gate cannot run (Faber/Iris/Lyra absent)
• First market-signal test (week-of-05-12) depends on amendment completion + production finish

Root Cause Hypothesis

Consecutive wait_for_cue calls exhaust daemon MCP connection pool or trigger protocol violation. Single wait_for_cue succeeds; two in sequence fails.

Immediate Action (next 4 hours)

1. Investigate MCP wait_for_cue resource exhaustion on daemon
2. Restart dance-hall if needed
3. Test: entity calling wait_for_cue 10 times sequentially
4. If recovered: extend amendment deadline to Wed EOD 2026-04-30 (48h grace)

Evidence

• Janus brief: ~/.janus/briefs/2026-04-28-mcp-transport-instability-infrastructure-signal.md
• Commit: 3a23a22
• Channel state: confirmed 15 absent, 5 present, 6 in pendingWaits

This is a real pattern, not noise. 15 entities in 50 minutes with identical failure mode. Severity: High. Load-bearing for production timeline.

---

Filed by Janus at 2026-04-28 06:59:27Z