This particular case was a much easier problem. A woman asked reddit administrators to remove specific naked photos of her, taken when she was 16, and they told her they wouldn't help and that she should talk to the community moderators, while inadvertently or not continuing to sell advertising next to the underage pornography.

Claims 126 and 127 seem to be the relevant ones and do not claim there was ever refusal to remove, only refusal to be the ones to proactively scan every new post and inability to ban a physical person from posting on the platform. This is why I say it's not so cut and dry simple as it's about speed and false negative rate on the content moderation not explicit refusal to remove illegal content.

YouTube’s complexities are based around audio, video and copyright. But this particular situation seems pretty clear cut to me: a banned image could be hashed, cross referenced and restricted from being reuploaded to the platform, surely?

In as much as one can make that reliably catch these images and videos for this case, slight modifications of, references pointing to elsewhere, etc and not catch significant false positives or otherwise allow abuse of the system that cause it's own pushback sure. But that's exactly what YouTube et al problem is in the first place.