What is people's experience with the on-board monitoring (prometheus)? I find that it consumes too many resources and destabilizes the clusters, so I just turn the dang thing off. Devs like to start it up and maybe look at the pretty graphs one in a blue moon. Today I'm looking at memory consumption and just want to flip the dang thing off again.

have you tried to change default polling interval? I believe default now is 30s