fix(evals): 300s budget floor for the two 90s design-consultation SDK tests

Root cause of PR #2533's e2e-design CI failure: design-consultation-preview
failed 3 attempts at 0 turns/$0.00/93s — the session was up but the model's
first completion queued past the 90s inner budget under concurrent API load
(11 matrix jobs; the sibling research test booted its first tool at 4s, so
this is API-side queuing, not CPU boot contention). The test was selected
only because touchfiles.ts is a global touchfile; the tested behavior is
untouched by this branch.

90s budgets cannot absorb one slow first completion. Both 90s tests in the
file move to the repo's saturated-runner standard (300s inner / 360s outer,
matching review-dashboard-via and retro-base-branch). Deliberately NOT
re-arming the runner's inner timer on first stream event: an audit found
~100 outer bun-timeout literals sized inner+30-60s that a re-arm would
silently break — the structural options are written up in TODOS.md.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

TODOS.md

@@ -233,6 +233,31 @@ SKILL.md untouched). `bun test` is green again.
 
 ## Scope-gate follow-ups (filed via /plan-eng-review on the plan-mode auto-select-B change)
 
+### P2: SDK eval budgets charge API-queue latency to the work budget — pick a structural fix
+
+**What:** `runSkillTest`'s single `setTimeout(timeout)` arms at spawn, so session
+startup AND the model's first-completion queue time are charged against the
+test's work budget. Under concurrent load (11 CI matrix jobs, or local eval
+runs sharing the org API), a first completion can queue 60-90s+, producing the
+deterministic `0 turns / $0.00 / <budget>s x3 attempts` failure shape. Observed:
+`review-dashboard-via` (PR #2472, 180s→300s), `retro-base-branch` (240s→360s),
+`plan-ceo-plan-mode` (300s→420s, 2026-08-12), `design-consultation-preview`
+(90s→300s, PR #2533 CI). Every fix so far is a per-test budget bump.
+
+**Why not just re-arm the timer on first stream event:** an audit (2026-08-12)
+found ~100 outer bun-timeout literals sized as inner+30-60s; re-arming the inner
+clock breaks every outer/inner relationship and needs a codemod of all of them.
+
+**Options:** (a) two-phase timer in session-runner (startup grace, re-arm on
+first NDJSON line) + codemod outer literals to inner+grace+slack; (b) adopt a
+300s floor for all CI SDK budgets (statically enforceable — a free test can
+assert no `timeout: <300_000` in skill-e2e files) and stop re-litigating per
+test; (c) startup-spawn semaphore in the runner (bounds the boot stampede but
+not API-side queuing — evidence says queuing dominates, so likely insufficient
+alone). Recommend (b) short-term + (a) properly sequenced with the codemod.
+
+**Depends on / blocked by:** none.
+
 ### P2: Wire the four demoted plan-mode/finding-floor PTY tests into periodic CI
 
 **What:** `evals-periodic.yml` runs an explicit 9-file matrix; the four tests

test/skill-e2e-design.test.ts

@@ -176,7 +176,13 @@ Include: color trends, typography patterns, and layout conventions you observed.
 Do NOT generate a full DESIGN.md — just research notes.`,
       workingDirectory: researchDir,
       maxTurns: 8,
-      timeout: 90_000,
+      // 300s, not 90s: saturated-runner class (same as review-dashboard-via /
+      // retro-base-branch). PR #2533 CI observed the sibling preview test at
+      // 0 turns/$0.00 for 93s x3 attempts — session up, first completion
+      // queued past the budget under concurrent API load. 90s budgets cannot
+      // absorb one slow first completion; 300s is the repo's standard floor
+      // for CI SDK tests. Outer timeout below rises to 360s for headroom.
+      timeout: 300_000,
       testName: 'design-consultation-research',
       runId,
     });
@@ -206,7 +212,7 @@ Do NOT generate a full DESIGN.md — just research notes.`,
     }
 
     try { fs.rmSync(researchDir, { recursive: true, force: true }); } catch {}
-  }, 120_000);
+  }, 360_000);
 
   testConcurrentIfSelected('design-consultation-existing', async () => {
     // Pre-create a minimal DESIGN.md (independent of core test)
@@ -274,7 +280,9 @@ Write a single HTML file to ${previewDir}/design-preview.html that shows:
 Do NOT write DESIGN.md — only the preview HTML.`,
       workingDirectory: previewDir,
       maxTurns: 8,
-      timeout: 90_000,
+      // 300s, not 90s: this is the test that failed 3x at 0 turns/$0.00/93s
+      // on PR #2533 CI — see the research test's comment for the class.
+      timeout: 300_000,
       testName: 'design-consultation-preview',
       runId,
     });
@@ -303,7 +311,7 @@ Do NOT write DESIGN.md — only the preview HTML.`,
     }
 
     try { fs.rmSync(previewDir, { recursive: true, force: true }); } catch {}
-  }, 120_000);
+  }, 360_000);
 });
 
 // --- Plan Design Review E2E (plan-mode) ---