Internally we have a notion of operation "deadlines." For example, when an operation is run with a maxTimeMS, we use this mechanism.

On operations with multiple batches there's some complexity here since we "roll over" the unused time from each operation to the subsequent one.

We should consider logging an operation's deadline, if it has one, in slow query logs. This should be easily comparable to the operation's end time (which is already logged), and will let us see how much longer the operation took compared to its budget.