[yaacovcorcos]

What this changes

The time limit for the three Test Web jobs in .github/workflows/ci.yml goes from 10 to 15 minutes. It's marked as a Scient change, since upstream uses 10.

Why

The slowest web shard now takes about 10 minutes, setup and cleanup included (9 min 50 s to 10 min 18 s on recent runs). With a 10-minute limit, a shard can pass every step and still be cancelled during cleanup.

On #634's merge-queue run, Test Web 1 passed all its tests but was cancelled at 10 min 3 s. The queue counted the cancellation as a failed check and ejected the PR.

Times from that run:

Shard	Setup and cleanup	Tests	Total
1	about 1.4 min	8.3 min (+18 s editor timing checks)	about 10 min
2	about 1.2 min	5.3 min	about 6.5 min
3	about 1.2 min	8.2 min	about 9.5 min

15 minutes keeps roughly 50% headroom over the slowest observed shard, without padding far beyond it. The server shards already follow the same rule: their limit is 35 minutes, set because "a green shard gets cancelled mid-cleanup" otherwise.

Not done here: splitting the web tests into more shards would make them faster. But the shard names are required checks in branch protection, and the queue-reuse script expects three shards, so that's a separate decision.

🤖 Generated with Claude Code

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.6 KiB	5.6 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	4.3 KiB	4.3 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.3 KiB	1.3 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	26.2 KiB	26.2 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.6 KiB	5.6 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	4.3 KiB	4.3 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.3 KiB	1.3 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	26.6 KiB	26.6 KiB	0 B (0.0%)	29.3 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: d5c9445 · PR result: ba4c80c · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.
Full compact HTTP snapshot with live cursor.

• Codex decoded thread snapshot: 133.1 KiB
• Claude decoded thread snapshot: 133.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.