In BF-42095 the stepdown_fassert_on_rstl_timeout.js test times out while waiting on ReplSetTest.waitForState() to return.

The theory is that when we do the hello() command on the crashed node, the connection was not properly closed by the server, and we hang until the TCP timeout is breached, which can be longer than the test timeout. See a more detailed explanation here.

To fix this we should retry to connect to the node when checking if it's down.