* it’s purpose built for mega-sized monorepo models like Google (the same company that created it)

* it’s not at all beginner friendly, it’s complex mishmash of three separate constructs in their own right (build files, workspace setup, starlark), which makes it slow to ramp new engineers on.

* even simple projects require a ton of setup

* requires dedicated remote cache to be performant, which is also not trivial to configure

* requires deep bazel knowledge to troubleshoot through its verbose unclear error logs.

Because of all that, it’s extremely painful to use for anything small/medium in scale.