Rust Ownership vs Python Garbage Collection — Which approach wins?

I've been building services in both languages and want to discuss ownership semantics versus GC tradeoffs. Rust's borrow checker catches data races at compile time, which is huge for systems work but adds a significant learning curve. Python handles everything with reference counting plus a cyclic garbage collector, which simplifies development tremendously while introducing unpredictable pause times under heavy load. For high-performance services, I've found that the ownership model prevents many bugs before they reach production. However, the developer velocity gap is real — things that take an afternoon in Python can take days to get through the borrow checker. What are your experiences with performance versus productivity tradeoffs between these two memory management models?