If the seedlist item contains the primary and there is a non-reachable member that times out it can cause the initially connected primary to timeout before the discovery processes has finished leading to a "no primary found".

Fix is to make initial topology sweep not start up the SDAM monitoring but to schedule it in parallel before the initial discovery is performed. This will make the monitoring thread keep the discovered servers alive and negate the "no primary" found.

https://github.com/christkv/mongodb-core/commit/a8918ac99c0325aeccb1e729704d0b0c7b38a3ff