It uses also tracing GC to handle RC loops.

I guess doing it this way allows to have mostly deterministic collection (that is then just an implementation feature, not language semantics) while needing the actual tracing GC to not be that efficient, as it presumably needs to be executed less rarely.

Ref counting is a GC-algorithm, and the slowest at that (especially with multi-core sync, but that’s not the case with python), as a tradeoff for less memory consumption.