How many agents can 2×4090 actually run at once? Three weeks of llama.cpp concurrency data — soft cap 5 @ 64k, hard cap 9, and why.

I'm the CTO of a mid-size nonprofit, and I run a local Qwen stack as a coding subagent for that work — partly on principle (a fair amount of what I handle shouldn't leave the building), partly because the token bill for bulk code work adds up fast on a nonprofit budget. Over three weeks I benchmarke