We are running a mongodb v4.0 replSet on bare metal servers. Recently one replSet member crashed and all data was lost. To bring it back into the replSet, all databases had to be cloned. A few days later, the replSet member was up & runing & in sync.
However, we observed that the size of the _id index of a (capped) collection was more than twice as large (423 vs 180 GB). The index sizes of the other secondary are quite similar to the primary though.
That's a problem because such huge _id index does not fit in RAM anymore since the server has "only" 376 GB of RAM.

Please see attached the collection stats of the primary and both secondaries.