[Bug]: Subagent thread bottom bar (ProviderSubagentBar) displays parent thread's model and effort instead of actual subagent configuration #17694

Before submitting

• I searched existing issues and did not find an exact duplicate. (Follow-up to [Bug]: V2 Codex subagent model stays "Not reported" after metadata lookup fails #15250 and [Bug]: Orchestrator V2 shows no reasoning effort for subagents #15214, specifically focusing on the client UI presentation in ProviderSubagentBar).
• I included enough detail to reproduce or investigate the problem.

Area

apps/web

Steps to reproduce

1. Start a parent thread using GPT-6.1-Sol with Extra High reasoning effort (gpt-6.1-sol / xhigh).
2. Spawn a native Codex subagent specifying gpt-6-luna with max reasoning effort.
3. Wait for the subagent to complete its run.
4. Open the dedicated thread view of the spawned subagent in T3 Code.
5. Inspect the bottom status bar (ProviderSubagentBar).

Expected behavior

The bottom status bar of the subagent view should show the actual model and reasoning effort configured for and executing the subagent:
GPT-6-Luna Max Completed in 1m 15s Runs on its own

Actual behavior

The status bar displays the parent thread's model and effort instead:
GPT-6.1-Sol Extra High Completed in 1m 15s Runs on its own

Investigation & Root Cause

Inspecting the runtime state and databases reveals the breakdown:

1. Codex execution truth:
   • In ~/.codex/state_5.sqlite and the rollout .jsonl session file:

     { "model": "gpt-6-luna", "reasoning_effort": "max", "thread_source": "subagent" }

2. T3 projection state:
   • In orchestration_v2_projection_provider_threads:
     nativeMetadata is null.
   • In orchestration_v2_projection_threads:
     modelSelection was cloned verbatim from the parent thread:

     "modelSelection": {
       "instanceId": "codex",
       "model": "gpt-6.1-sol",
       "options": [{"id": "reasoningEffort", "value": "xhigh"}, {"id": "serviceTier", "value": "default"}]
     }

3. UI rendering logic (ProviderSubagentBar.tsx):
   • In apps/web/src/components/chat/ProviderSubagentBar.tsx:
     When nativeMetadata.modelSelection is null (via deriveReportedModelSelection), the UI falls back to props.selectedThread.modelSelection.model and formatModelSelectionEffort(props.selectedThread.modelSelection).
   • As a result, the parent's configuration is presented as the child's, misleading users into believing the subagent consumed the parent's flagship tier.

Impact

Minor bug or occasional failure. Causes confusion regarding actual model consumption and token expenditure across subagent runs.

Version or commit

T3 Code Nightly 0.0.46-nightly.20261010.2908

Environment

macOS 26.7, Apple Silicon, T3 Code Desktop, Codex CLI 0.162.1

Related Issues

• [Bug]: V2 Codex subagent model stays "Not reported" after metadata lookup fails #15250 (tracks the initial empty-rollout error during server metadata lookup; note that PR fix(codex): retry child model lookup after empty rollout #15285 is scoped to server-side metadata recovery only)
• [Bug]: Orchestrator V2 shows no reasoning effort for subagents #15214 (tracks missing reasoning effort in V2 subagent contracts)