From their own website:

"Accurate Multi-Level Verification Use swift and accurate multi-level cyber verification techniques to confirm copyright infringement"

It sure doesn't sound to me like a company that would like it on record that they aren't as good as they think they are. Have them publicly admit the cock up.

https://mxtoolbox.com/SuperTool.aspx?action=dmarc%3aopsecsec...

But they are getting their own forensics feeds.

This seems to me that they are willifully staying in p=none so they can try and have plausible deniability should these things get public.

"Here's the hash of the file you're infringing." I mean, where did they get that from? It's a rather specific thing for a very specific file from a very specific source. How can that possibly be a mistake? At best it's a result of some automation that indiscriminately makes these claims, which doesn't fall under "cock up" in my book.

at about the same time there was a practice of a running padfile, that you would ammend every time you retorrented so that the hash would change and make it hard to kludge out the distribution network by tracking known hashes.

in summary it think it is very possible to "forge" hash numbers and impersonate or obfusicate the content

hash collision is rare, intentional hash mimicry maybe is not so rare.