The WiredTiger pull request testing currently only runs test/format for 10 minutes on x86. That isn't enough to give robust signal about fallout from a change.

We experienced that failure with WT-12435, which trigger 58 test failures within hours of being merged.

That particular failure mode wasn't detected because we did not include ARM test/format testing - and the issue was ARM specific. It also makes sense to do more stress testing on each PR push.

I was hoping this would be a borderline trivial change to test/evergreen.yml, but I think it's a bit more complex. It will require adding an Ubuntu 20.04 ARM variant into PR testing, and ensuring it has the format-stress-pull-request-test task included.