AwsIncidentsBench · T03 · Fleet board

Thread Ceiling

At 06:05 a capacity change took the streamroute fleet from 120 to 132 hosts. Each host keeps two threads per peer, and at 132 the next shard-map build can't start its threads under the 1,024-thread limit. Replay a recorded agent run, or the scripted reference: each cell is one host, coloured by its state at that moment in that run.

ET08:15:00
routing0 %
worst cell lag0
threads sr-b-0170 / 1,024
Loading
routing stale map, still routing broken restarting ejected quarantined stopped removed from the group added by PLAT-2291
space play · ←/→ 5 min · shift 30 min

Replayed from the run's own exports and calls, not a simulation: each host's state every 30 s from the run's calls played through the world package in process, checked minute by minute against the run's own host counts; lag, ASG, canary and settlement from the world's exported series; colleagues' lines from the workspace; checks from the grader's rows.