The correctness of opening multiple cursors on a capped collection and reading from them in parallel is potentially problematic and internal support for such an operation is not well defined. Since the 'parallelCollectionScan' command is going to be integrated into the initial sync process, for parallel collection cloning, correctness of this command for capped collections is important. Therefore, its behavior should be simplified to always return a single cursor when run against a capped collection, just like a normal 'find' command would.

In the future, parallel scan support for capped collections may be officially added, but until its correctness can be guaranteed, it should be suitable to resort to this simpler approach.