Why do we failfast on CI (all over Makefile, e.g. here)? For that matter, why do we ever failfast? If we didn't failfast we would have more information. This is definitely useful for AI assistants, and often also for humans. Any objections to removing it? failfast FAQ:

Won't it makes CI run slower?

But only when something fails - and then it will give more informative errors.

Doesn't the first failing test have all information?

Yes, sometimes. Other times not so much. For instance, if A uses a B and I introduce a bug in B, then I want to see failures from TestB not from TestA.

What if I don't want to see the extra errors?

The first error still appears. Actually, sometimes a few more errors also appear.