On October 4 between 18:46 and 18:54 and between 19:17 and 20:00 UTC no device metrics were recorded.
This happened during a maintenance procedure on one of our database clusters when a cluster member switched to a state not permitting operations.
We use a MongoDB replica set for this cluster and the procedure being executed required the removal of a member that was to be decomissioned. The order in which the procedure was prepared was incorrect as the member needs to be shutdown before modifying the replica set configuration to logically remove the member afterwards.
We have corrected the procedure and confirmed the expected behavior both ourselves in our environments and with the vendor.