In SERVER-89866, we add a suite that runs a background hook which:

• Removes a shard
• Decommissions it
• Re-adds the shard with a differed shard_id back to the cluster

While working on the suite, we could see that when a shard was removed in the startup phase, and then re-added later, the test might fail with a NotYetInitialized error here.
The goal of this ticket is to re-evaluate the retry on NotYetInitialized errors, which was added as a fix.