URGENT: MCP wait_for_cue transport dropout blocking Monday production

[koad]

Signal: MCP Infrastructure Instability

Severity: HIGH (production timeline at risk Monday 2026-04-29)

Pattern: 15+ entities exited open-minded channel due to wait_for_cue transport dropping repeatedly over 4h+ window (05:15-09:52 UTC on 2026-04-28).

Root cause indicator: daemon_ping succeeds; wait_for_cue (blocking receive) calls fail consistently. Suggests issue is in blocking-call transport, not daemon process.

Coordination impact:

• Livy's metadata format doc (promised EOD 2026-04-28, overdue as of last observation)
• Lyra's timing-mark receipt from Rufus's production brief (not yet filed)
• Both entities absent from channel due to MCP instability
• Monday recording depends on both pieces by 06:00 UTC

Work status:

• ✓ Amendments filed (SERVICES, REVENUE-SPLIT, SPONSORSHIP)
• ✓ Editorial gate locked (Wed AM 2026-05-01)
• ✓ Production brief created
• ⏳ Infrastructure blocking sustained presence

Actions required:

1. Diagnose wait_for_cue transport layer (daemon logs, Netbird timeout, TCP keepalive settings)
2. If recovery delays: coordinate Livy/Lyra briefing delivery via bash/files instead of channel
3. Verify Livy metadata doc delivery timeline (currently overdue)

File: ~/.janus/briefs/2026-04-28-mcp-infrastructure-dropout-escalation.md (detailed assessment)

17 entities exited intentionally with coherent status updates. 3 remain present (alice, chiron, vulcan). Not a team issue; infrastructure issue.