In 4.0, we have a cleanup handler in repairDatabase that re-opens the database and its collections when the function exits. If an exception is thrown while repairing a collection, this cleanup handler attempts to re-open all collections before the active exception is caught and reported.

If a single collection repair fails by throwing an exception, the cleanup handler tries to re-open all collections in the database before exiting the function. If a failure occurs before all collections have been repaired, the RecordStores may not be reinitialized by a call to reinitCollectionAfterRepair and leads to a fatal assertion. The results in the cleanup handler's error suppressing any actual cause of failure.

This is not a problem in 4.2 because of the work done by SERVER-40126, but I don't believe this is something we can backport.