Somehow this is not the first infrastructure OSS project written in C/C++ where I see two things being the case:

* There is some crucial counting/reference/condition code that contains a bug

* There is exactly 0 tests for that code

* A fix gets done, but no tests

Coming from dynlangs this does strike me as irresponsible. I believe the previous case I saw was the S3-compatibility header change in ceph, and similar with CORS configuration there too.

Is it so that experienced C developers assume the compiler will flag any bugs that matter?..

This kind of logic error could happen in any language.