The reshardingChangeStreamsMonitor's callback creates an operationContext from the 
cancelableOpCtxFactory to update the state document. However, this call can race with the main run() thread of the donor here if abort was called.
 
This issue isn't very common since the reshardingChangeSteamMonitor does take in a cancelToken from the donor/recipient and should interrupt its work if the cancelToken is cancelled. One solution would be to capture the factory object in the callback here.