Ah, that’s because we use

--lint-per-target-caching

there because we have remote caching so it results in much better performance. A followup project I have is to get the best of both

--no-lint-per-target-caching

(generally faster perf due to the very high overhead of linters + less verbose output)and

--lint-per-target-caching

(fine-grained caching)