[jeffdupont]

TestProcess::test_boot_is_outside_the_caller_timeout gave the stream a 0.2 s timeout after a 0.6 s fake boot. On a slow runner the stream alone can take more than 0.2 s, so the test failed with no bug present (it failed CI on #201).

This raises the fake boot to 2.0 s and the timeout to 1.0 s:

• The boot is still longer than the timeout, so the test still fails if the boot moves inside the caller's timeout. I checked this locally by wrapping async with handler in the timeout: the test fails.
• The stream now has 1 s of room instead of 0.2 s.

The test takes about 2.5 s instead of 0.6 s.

🤖 Generated with Claude Code

---

Note

Overview
Relaxes timing in test_boot_is_outside_the_caller_timeout so it stops failing on slow CI when the stream alone exceeds the old 0.2 s budget.

The fake OpenCode delay_ready goes from 0.6 s → 2.0 s, and the asyncio.timeout around streaming goes from 0.2 s → 1.0 s. Boot is still longer than the timeout, so the test still catches boot work incorrectly running inside the caller’s timeout; the wider window only gives the stream headroom on sluggish runners. Comments in the test document that intent.

^{Reviewed by Cursor Bugbot for commit 53ea3d2. Bugbot is set up for automated code reviews on this repo. Configure here.}