Collect the following metrics at cursor level (currentOp) on mongoD:

• cursor.changeStreams.optime
  • Operation time (cluster time) of the current oplog entry at cursor position.
  • The optime of the latest event that we've sent or the current 'high-watermark'.

An important implementation decision is when to update the metrics, since some of the oplog scanning functionality is done by the classic find-subsystem. A viable option is to do this in the earliest change stream stage of the aggregation pipeline ($_internalChangeStreamOplogMatch).