It would be great if we could set an op_timeout per-operation (per-cursor, perhaps).

The vast majority of mongo calls we make should be quite fast (e.g. _id lookups), and so should have a short op timeout, but we occasionally need to make a longer query for an analytics call or something, and it would be great to be able to customize the op_timeout for those cases.

Right now, we have to set the op_timeout as long as the longest mongo query we will ever make, which means that it may take us minutes to detect a failure in the worst case, even if we're only issuing an _id lookup that should normally succeed in <1s.