In OplogFetcher, we use a different timeout on recreating the cursor. And after SERVER-57617, the oplog fetcher no longer uses a tailable cursor and we re-create a new cursor every time the cursor is exhausted. And this uses a 2s timeout because of this.

We could consider using a larger timeout in this case. But we should understand whether it is expected for the aggregation to exceed the 2s timeout.