Trusting builds with Bazel remote execution (opens in new tab)

Bazel solves a ton of problems other systems don't, but it introduces a ton of its own. If you can afford (or are currently accidentally affording) 2-3 FTEs managing builds, that's all you need to scale Bazel indefinitely and solve basically all of the rest of your build problems. If you can't afford a dedicated Bazel team then it's absolutely not worth using on anything more than a single "team" (people talking to each other every day, a dedicated Bazel xoogler, probably 5-13 people).

As to the actual "why", even little things like BUILD files requiring you to enumerate dependency graphs gives you a leg up that you _can_ implement in other build systems, but likely won't. Whether you're a "monorepo" or not, you still have all your code living _somewhere_ as a monorepo, and the only question is how good your tooling is. How easy is it in your system to "run all tests properly depending on the set of changed files"? That's easy in Bazel and hard in every other solution I've seen (possible of course, but those sorts of constraints aren't the happy path for other build systems, so teams don't tend to build that way without a very solid lead imposing that strategy).

Bazel is complex, yes, and in my opinion not suitable for small, one language projects.

But I’ve seen many builds, outside of FAANG, that are too slow and that also break frequently after a simple “git pull”. Which other systems promise to improve those, and how do they do it?

Mods sometimes merge posts. Whether any particular merger makes sense can be a bit of a toss-up.

At the very least, bazel would devour the cmake responsibilities in this case. That's not intended as a criticism of either. Just a description of bazel. It can feel very Star Trek Borg like.