1.) Simple alteration (change a pixel in MS paint) or encryption of content bypasses the filter 2.) Patching out the filtering routine bypasses the filter 3.) Blocking the phone-home address (pihole, router firewall, etc) bypasses any reporting 4.) Any vulnerability in the future that allows an attacker to report arbitrary clients (disclosure of client IDs, weakness in app, weakness in server) renders evidence gathered by the system unreliable.
At best clientside filtering allows you to draw relationship maps of technically incompetent perverts who might possibly be sharing CP. What harm reduction are they trying to get out of that?? Why not just refocus efforts on catching the small minority of individuals who are actually producing this content??
But hey, if these garbage clientside filtering of image uploads is enough security theatre to keep governments satisfied, I say let them have it.