I'm working on multiple tools to make concrete programming language better.
leanprof is one of them. leanprof profiles Lean programs and proofs without changing your project. It counts what makes Lean slow that no other profiler sees (copies of shared arrays and strings, allocations, failed reuse, the unfolds and instance searches behind compile time), names the line and the cause, and prints the fix. In CI it fails a pull request on "+1 copy" and says what made it slower.