I don't think we ever got many direct complaints about GPL, but our main assumption was that copy left contamination was pushing developers away. We wanted more open licensing to encourage wider adoption, so we started a replacement project under MIT license, I believe.
Once the replacement project was in a good state, suddenly plugin developers started crawling out of the woodwork.
My estimation is that about 30% of our trouble was due to licensing concerns, and the rest was down to the original project being in a very, very bad way. It was very old, had been abandoned by six developers, and about 70% of the code was unused and unusable.
The new project had bells and whistles and was more attractive, but I think the open license definitely helped.
You probably could have built a plugin without needing to use the GPL, but it would have been a lot more difficult. That whole project was a disaster, and I never once regretted letting it finally die.
Cygwin changing to the LGPL back in 2016 enabled me to distribute BSD code with Cygwin DLL's without having to GPL the code. When I learned from the Cygwin mailing list about this change, I immediately set to work to create a fork of the Cygwin DLL which provides more native-like behaviors, to be used as run-time library for programs. Had they not announced a licensing change, I'd not have done anything of the sort.
https://librearts.org/2012/01/whats-up-with-dwg-adoption-in-...
I'm the maintainer, the FSF owns the code, and I'm very fine that it's GPLv3. Others not so, but we don't care. GPLv3 is the only possibility.
If "this product is made from a combination of domestic and foreign components" is good enough for physical goods, then it should be good enough for software.
But IANAL. :)