I'd like to see more interview questions that force you to explain a trade-off instead of just define a Python concept.

For example, imagine a FastAPI service that processes thousands of incoming requests, calls several external APIs, and then runs CPU-heavy feature extraction before returning a response. Switching everything to async def won't make the CPU-bound part run in parallel, and adding a thread pool might just move the bottleneck. You'd need to separate the I/O-bound work from the CPU-bound work, measure where the time is going, and decide whether threads, processes, or a different execution model actually fits.

That's the kind of question that tells me more about a developer than asking them to recite what the GIL does. Can they explain the cost of moving data between processes, handle timeouts and cancellation, and back their choice with profiling results?

Knowing the API is one thing. Knowing when it stops being the right tool is where the interesting engineering decisions start.