pr #37714 runtime verification baseline github/main eebb6bccdf, isolated /home/user/main/.t3 real codex gpt-6.1-sol replied BASELINE_READY through the web client. $ t3 thread send "reply" --model gpt-6.1-sol --effort high main: Unrecognized flag: --model; Unrecognized flag: --effort original pr head 8f81f0174078878098656a78d72f8b87ea600a51 same CLI send on a real Codex thread: Model gpt-6.1-sol on provider codex rejected options: Unknown option effort; supported options: reasoningEffort, serviceTier. fixed head 50351e5fa1 $ t3 thread send "Reply exactly SWITCHED_SOL_HIGH. Do not use tools or edit files." --model gpt-6.1-sol --effort high --json {"threadId":"1dcb0e60-f6e2-4cad-8289-6fa24e113ea8","link":"https://app.t3.codes/threads/88a26f2c-2c6f-4b6e-a23d-a28c38ca3d67/1dcb0e60-f6e2-4cad-8289-6fa24e113ea8","messageId":"message:mcp:client%3A65177dfc-3870-4654-811a-c645883b7bea:thread-send:262d0beb-98d7-4b5d-9c90-d1ac2db61e94","runId":"run:thread:1dcb0e60-f6e2-4cad-8289-6fa24e113ea8:ordinal:3","status":"starting","delivery":"started","modelSelection":{"model":"gpt-6.1-sol","options":[{"id":"reasoningEffort","value":"high"}]}} real provider replied SWITCHED_SOL_HIGH; client changed from Astra to Sol/high. busy-turn send (modelSelection, omitted mode) through authenticated MCP: { "threadId": "1dcb0e60-f6e2-4cad-8289-6fa24e113ea8", "link": "https://app.t3.codes/threads/88a26f2c-2c6f-4b6e-a23d-a28c38ca3d67/1dcb0e60-f6e2-4cad-8289-6fa24e113ea8", "messageId": "message:mcp:client%3Aaf9df1a7-5996-442d-b107-d63907c10f37:thread-send:verify-busy-queue", "runId": "run:thread:1dcb0e60-f6e2-4cad-8289-6fa24e113ea8:ordinal:7", "status": "queued", "delivery": "queued", "modelSelection": { "model": "gpt-6.1-sol", "options": [ { "id": "reasoningEffort", "value": "high" } ] } } web client showed queued input while Sol/low ran timeout 15 sleep 8. BUSY_COMPLETE completed before QUEUED_HIGH_COMPLETE began at Sol/high. read-only persisted run check: run 6 completed 13:28:03.732Z, run 7 started 13:28:04.150Z. neighboring checks: invalid model { "content": [ { "type": "text", "text": "{\"_tag\":\"OrchestratorMcpFailure\",\"code\":\"model_unavailable\",\"message\":\"Model gpt-no-such-model is not advertised by provider codex. Advertised models: gpt-6.1-sol, gpt-6-astra, gpt-6-luna, gpt-6-sol, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, gpt-5.5.\"}" } ], "isError": true } invalid effort { "content": [ { "type": "text", "text": "{\"_tag\":\"OrchestratorMcpFailure\",\"code\":\"invalid_request\",\"message\":\"Model gpt-6.1-sol on provider codex rejected options: Option reasoningEffort must be one of: low, medium, high, xhigh, max, ultra.\"}" } ], "isError": true } refuse steer { "content": [ { "type": "text", "text": "{\"_tag\":\"OrchestratorMcpFailure\",\"code\":\"invalid_request\",\"message\":\"modelSelection needs a send that starts or queues a turn, not mode:'steer'.\"}" } ], "isError": true } refuse restart { "content": [ { "type": "text", "text": "{\"_tag\":\"OrchestratorMcpFailure\",\"code\":\"invalid_request\",\"message\":\"modelSelection needs a send that starts or queues a turn, not mode:'restart'.\"}" } ], "isError": true } receipt replay { "threadId": "1dcb0e60-f6e2-4cad-8289-6fa24e113ea8", "link": "https://app.t3.codes/threads/88a26f2c-2c6f-4b6e-a23d-a28c38ca3d67/1dcb0e60-f6e2-4cad-8289-6fa24e113ea8", "messageId": "message:mcp:client%3Aaf9df1a7-5996-442d-b107-d63907c10f37:thread-send:verify-auto-low", "runId": "run:thread:1dcb0e60-f6e2-4cad-8289-6fa24e113ea8:ordinal:4", "status": "completed", "delivery": "started", "modelSelection": { "model": "gpt-6.1-sol", "options": [ { "id": "reasoningEffort", "value": "low" } ] } } plain follow-up: PLAIN_RETAINS_LOW completed, model remained Sol/low. reinstalled with vp i, stopped captured dev process group, restarted same isolated home. plain send after restart: {"threadId":"1dcb0e60-f6e2-4cad-8289-6fa24e113ea8","runId":"run:thread:1dcb0e60-f6e2-4cad-8289-6fa24e113ea8:ordinal:9","status":"completed","reply":"RESTART_PERSISTED_HIGH","error":null} CLI send with explicit --model gpt-6.1-sol --effort high after restart: {"threadId":"1dcb0e60-f6e2-4cad-8289-6fa24e113ea8","runId":"run:thread:1dcb0e60-f6e2-4cad-8289-6fa24e113ea8:ordinal:10","status":"completed","reply":"RESTART_CLI_HIGH","error":null} web picker remained GPT-6.1-Sol, High after reload. blacksmith: vp run check:changed HEAD passed format, lint, knip, GraphQL, 3 package typechecks, 18 tests. HEAD is required because the testbox has an orphan source snapshot; comparison covers the two fix files. blacksmith focused PR checks: service 97 tests + contract 3 tests passed. existing GitHub agent-context-budget check fails: 52145 -> 52721 (+576), label changes prohibited.