Most of the stories I've read about terrorists and mass shooters report that they were using bog standard instant messaging, unencrypted e-mail, and social media. Most of these people are not highly technical and do not practice good opsec. Hell Dread Pirate Roberts was pretty technical and still got busted because of bad/lazy opsec, not (as far as anyone knows) because encryption or Tor were broken.

The child porn thing is a bad faith argument too. Child sexual abuse is under-investigated and under-prosecuted already even when the information is there or when actual reports are made. (Adult rape is under-prosecuted too.) They don't do enough to go after child predators using existing tools, so why would more tools matter?

Large amounts of child porn detected equals platform being used on similar scale for producing it / distributing new content / grooming. By playing with equivalence, you can push for tougher policy.

It is under-prosecuted for frankly embarrassing reasons. Tech companies can't submit it on the spot. They have to wait for them to come to them, and have to delete it if they take too long.