In SERVER-80135, an existing failure in this test was made more frequent. This failure can be seen in this BFG which is closed (I think erroneously) as part of an existing build failure.

The problem seems to be related to a forced refresh happening on the primary triggered by one of the secondaries, though this hasn't been confirmed.

Once the issue in the above BFG is understood and addressed, we should re-enable the test and consider increasing the number of nodes in the shards again.