Debug OpenClaw model cascade failures causing Slack lag or unresponsiveness.
Use when: (1) Slack messages are slow but console/Claude Code works fine,
(2) logs show "No API key found for provider" or rate limit errors,
(3) "allow-always is unavailable" error when approving exec commands,
(4) model fallback errors like "All models failed" in gateway.err.log,
(5) the gateway PROCESS is up but every agent run dies with FailoverError
because all cascade providers are down at once (codex cooldown + paid
fallbacks out of credit), (6) you want to build a FREE OpenRouter fallback
basket so codex cooldowns stop killing the agent, (7) a free model fails with
"Upstream error from Op…
