I have followed your very helpful suggestion below, and restored the
mingw32 build in a virtualbox.
As 64bit seems required nowadays, I've gotten cygwin64 ready as well,
all up to well known and documented fork() failures which I have been
unable to resolve up to now (after trying all FAQ's, etc.)
1) Is it not perhaps the path of least resistance for a windows user to
simply run a virtual linux box for maxima? mingw and cygwin both have
multiple stability problems.
I think for a typical maxima use on windows, that's way too hard. It's bad enough to have to get VirtualBox. But then you have to figure out what Linux distro to use, and then figure out how to install it with the right VirtualBox settings. After that, you've got to install the necessary tools to be able to run maxima. If you're using gcl or ecl to run maxima, you'll have to figure out how to install the C compiler and libraries to go with that. And gnuplot. And so on.
That's way too much for a typical user who wants to run maxima. As a developer who only has a Windows machine, perhaps that's not too bad.