Max Weinbach7/14/2026, 4:33:18 PM

The full-suite runner exited at the first failing file, so later
failures never ran and there was no end-of-run recap. Full runs now
continue through every file, collect failing test names from a per-file
JUnit report (reporter text is unreliable on Bun canary), and print all
failed files and tests at the bottom before exiting non-zero. Targeted
runs (bun run test -- <files>) are unchanged.

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

Parent2fb2ee0