acbcb208372ccee8cf9d84b23de71633a2ed413d
MoE at 95C with p50=13s latency — thermal throttling causing death spiral. Both slots stuck processing for 113s p95. Dense idle at 38C with 2 free slots. Reducing MoE to 1 slot forces heavy overflow to Dense, giving MoE thermal headroom. Heavy tier: MoE → Dense → VLM still valid — first heavy goes to MoE, second overflows to Dense.
Description
Syslog Operational Agent Harness — Nginx routing, Redis queue, circuit breaker, monitoring, Docker migration
822 KiB
Languages
Python
74%
HTML
24.7%
Shell
1%
Dockerfile
0.3%