Drivers which use the streaming hello protocol for server monitoring can consume significant server resources. By maintaining a single stream to a backing mongod and caching the result, monguard can reduce the overhead associated with handling these requests.

monguard must open a single persistent stream for hello commands to a backing mongod, and cache the hello results.

Monguard must update its cache when MongoDB Server advertises new topology information via the stream.

Clients which open streaming hello requests to monguard will receive cached hello results.
Clients which poll hello for replication topology will see cached results.

Monguard's hello responses must reflect the client's split horizon topology, which they opted into via SNI.