I'm running an application on n8n with only one worker and the default execution settings. I've recently started experiencing slowness issues due to some queued requests.

So, what is the best cost-benefit for scaling this process?

Add more workers, increase the requests limits for each worker, add more full instances of n8n, try to join the flows to reduce the number of requests at the same time, or some other solution?

Additionally, the worker is stopping at 2GB of memory, even though I have a plan with a 32GB limit and it's unlimited in the replica settings. This also appears to be causing slowness issues. Is there any way to increase/remove this limit?