I'm using p-limit for a bunch of batch processing, and it's been working quite well for some time.

Recently one of our services took advantage of our politeness and reduced their cluster size, and now I'm having trouble with finding a parallelism size which wakes up their service without overpowering it. In discussions someone commented that it would be nice if we could ramp up our parallelism over time.

I don't see anything in your suite of tools that allows that. I'm wondering if it needs to be more complicated than allowing the concurrency argument to be updated dynamically (I'm sure I could reach in and change it, but as it's not part of the public interface, who knows if/when you could break that cheat).

I'm picturing setting an initial limit of 10, and then setting 2 timeouts to ramp up to 25. Or perhaps n/2, n/1.5, n.