if the former, i’m curious where CPU time is being spent… if the latter, we’d need to figure out who is waiting for who, which might involve thread dumps