Possible race condition in new nodes joining the network causing ledger replication to not start

[gaurav137]

In test runs where in new nodes want join a network that is in Opening state without any read only snapshots to seed them are failing. Most likely it could be a race condition where both the joiner and the primary have opened TCP connections from both ends but neither makes forward progress after that.

File this bug to track the issue.

[achamayou]

@gaurav137 no luck on reproducing this so far, a set up that joins 15 nodes in quick succession, pre-open and without snapshot, always seems to converge fairly quickly (2s at most).