I am R&D'ing on currently leveraging Mongo ChangeStreams.

I am seeing that it has a good throughput, resilient to failures due to the 'Resume token', but I dont see any documentation on whether if/ how we can manage multiple instances of a subscriber application to receive changes in a load balance manner. Without this, we would have to use a single instance per filtered stream, inherently dealing with single point of failure.

If this is currently not offered by the mongo drivers (we use scala), it would be great help if you could direct us to some other implementations that could be done to achieve this.