Add cProfile to understand CPU bottlenecks

[leplatrem]

Note: locally, when loading all prod checks, I don't have bottleneck. Biggest time is IO with cold cache

That line again means your app is mostly idle in the event loop, not burning CPU.

<method 'control' of 'select.kqueue' objects> is macOS’s I/O selector (what asyncio and aiohttp use under the hood). It spends most of the time waiting for network sockets to be ready.

Seeing it dominate is normal for servers that are mostly waiting on I/O. It tells you:

Python threads are mostly sleeping, waiting for sockets to be readable/writable.

The CPU is not your bottleneck — requests are waiting on network, database, or external services.