One of the best things in bazel is that it's very straightforward when comes to tests - https://bazel.build/reference/test-encyclopedia#role-test-runner https:/

One of the best things in bazel is that it's very straightforward when comes to tests - https://bazel.build/reference/test-encyclopedia#role-test-runner https://bazel.build/reference/test-encyclopedia#role-test-ru... - you need to classify your test as short, moderate, long, etc., but also how much memory it takes. In addition if you tests requires isolation, put it too. So potentionally one can combine multiple independent code bases, that share the same testing framework (say gtest), into single executable, and distributevly execute them using sharding on multiple machines (e.g. depending on what changes were detected in the presbumit/preflight/whatever you call phase). And because the tests were marked as such and such, it'll use that knowledge to group them better, and if they no longer qualify, an error would be printed out, and how to fix this (if it's worth fixing).