Thanks, this is super informative! It sounds like maybe you weren't running multiple concurrent agents in the GPT-5.5 days, but now you are? Maybe you're on one of those US$200-a-month plans?
When working on trickier things that require me to test on device and report results, this would drop to only a single session. I can't context switch between multiple threads while focusing on an investigation where I need to steer the agent.
Yes, I'm using the $200 plan for Claude, and only $20 for Codex, where I use 6.1 Sol to review every change implemented by Opus 5.5.
In the past I paid for both $200 subscriptions, but with Claude's current usage limits, that is no longer needed.