I took a break since you had posted. Recently, I revised the map implementation to handle hundreds of files with lesser memory consumption. The llil4map variant may handle Chuma's use-case; see my reply to Chuma.

There is also the mcesort implementation, a GNU parallel (parsort variant). It includes a mini-MCE (less than 1,150 lines) integrated into the script. The application itself is around 500 lines. It runs on Unix only.