NODE-3893 Description
What problem are you facing?

We are on migration process of our Mongo servers to version 5.0. Making this happen, it was necessary an upgrade from our services to new driver (>=4.0) where we found a problem that didn't use to happen in a previous versions (<4.0).

We have been using the Mongo resource "collection watcher" for update our caches into the applications, but we noticed that Mongo Atlas makes a not planning rebalance of the nodes ReplicaSet. When this happen the class ChangeStream receives a close signal and the applications stop receiving new database updates.

What driver and relevant dependency versions are you using?

Driver NODEJS version 4.1

Steps to reproduce?

1 - Create a basic ChangeStream and listen the close event;
2 - Start application;
3 - Execute failover test into the interface Mongo Atlas, or change the primary node of the replicaSet manually;
4 - The application must receive close signal and stop receiving new updates.

OBS:
**

I am wait a solution for this problem some months, in the old versions(<4.0) this problem don't exists. It's rigth the solution of this problem to be a Improvement? 

This problem is forcing that my company take restart some applications always that mongo atlas changes your primary node. I believe that problem is a bug and should have more priority.

Thank you