Not an attack on the GNU GPL
It's worth pointing out that the author of shared-mime-info that raised the issue with mimemagic was open to modifications that avoided the need to relicence mimemagic (by using the XML from s-m-i at run-time rather than shipping it with mimemagic).
It was the mimemagic author that decided to relicence the code as GPL and then stop developing it, thus precipitating the forks, etc from folks looking for something that would be maintained. Including a fork that did precisely what the s-m-i author was willing to accept.
This situation arose purely because authors of code wanting the use the MIT licence (mimemagic wasn't the only project pinged over GPL violations of s-m-i) incorporated GPL source. And downstream users like RoR relied on the published licence and didn't audit the code. It's possible automated licence checking would have failed since the published XML in s-m-i's repo lacked the GPL notice due to a bug; that has been fixed.
If I wrote GPL'd code, I'd be happy for users to find alternatives with acceptable licences rather than violate my code's licence. A lot of the liberally licenced alternatives for MIME databases have significantly less useful matching data (i.e. the 'magic' contents part, beyond the file extensions); s-m-i considered one of the most complete and comprehensive available.