In SPM-4003, we introduced logic to avoid clearing the connection pool in response to connection establishment errors to ensure clients (mainly mongos) doesn't negatively react to rate limiting rejections from the remote host. However, the determination of which errors to do this for is quite coarse. The main case is HostUnreachable, but AsyncClient will convert pretty much any error it gets to this.

Instead, we should tag specific errors that are likely to be indicative of remote connection rate limiting and only skip processing the failure in that case. We can use the driver specification for guidance:
https://github.com/mongodb/specifications/blob/cadc0726cd5e3a201ede910e405b0731f6830cf5/source/connection-monitoring-and-pooling/connection-monitoring-and-pooling.md#connection-pool