@esk Are they part of a trace, and/or do they have a parent span?
I think they are Sidekiq workers polling Redis and trying to see if there are any jobs to process, in which case they can be filtered (but maybe not all ZSCANs?). If thats the case, we should probably filter them out directly in our instrumentation, or better sample them heavily.
For mastodon.social we only keep 1% of traces, except for errors where we keep 100% of them.