Verifying a downloaded file doesn't require a cryptographically secure hash function...

Of course it does, otherwise a malicious mirror can (theoretically) work to find a collision between their malware and the legitimate file and serve you the former.

There's no good reason not to use a secure hash function.

If your threat model involves an attacker who is able to achieve a hash collision while still implanting a sophisticated malware, you should probably avoid downloading software from random websites...

It would be pretty impressive, as they'd need their malware to both do what they want and exactly match that hash. Not impossible, just clever.

well it does require one that can't have collisions. otherwise what's the point in "verifying"?

There is no hash function that can't have collisions by definition.

*isn't known to have collisions, then

They have moved away from SHA-1, and are now using SHA-256. Previous releases were signed with SHA-1, and before that it was MD5: https://handbrake.fr/checksums.php

I'm not sure why they haven't retroactively calculated checksums for older versions.

But you're right that SHA-1 needs to stop being used: https://sites.google.com/site/itstheshappening/

These researchers have found the first "freestart" collision, and they estimate the SHA-1 collision cost to take a few months, costing between 75K$ and 120K$.

Practically speaking, I don't think anyone could make a profit by forging a Handbrake release, but the FBI probably do have some very high-profile targets who use video encoding software.