Why is my async Python HTTP request pool deadlocking after 500 concurrent connections?

I'm building a high-concurrency data scraper using `aiohttp` and `asyncio`, managing a fixed-size connection pool. After roughly 500 concurrent requests, the entire pool just locks up, no new requests go out, and existing ones never complete, without raising any explicit errors. What common `aiohttp` or `asyncio` patterns could be causing this silent deadlock, or are there specific OS-level limits I should be checking immediately?