The real question is why is that a problem? There are quite a few things involved in shadowing:
1) Processes retrieving/writing journal files
2) Dejournal reader process which reads those files and queues up records for sets/kills into database
3) Dejournal worker process which actually does those sets/kills
4) Dejournal prefetchers which fetch blocks from disk so 3) doesn't need to do disk reads itself
Most likely, you have 16+ dejournal prefetchers, optimizing the performance of shadowing keeping up with your source system.