Among other places, https://www.ma.utexas.edu/pipermail/maxima/2011/024573.html there was discussion about examples like the following (simplest found, due to Stavros, thanks!):
(%i3) desolve([diff(f(x),x)=f(x^2)],[f(x)]); 2 laplace(f(x ), x, g432) + f(0) (%o3) f(x) = ilt(------------------------------, g432, x) g432 (%i4)
There was some discussion at that thread, but no further followup, as far as I could find. Thanks!
I've pushed commit a3e1b463 which fixes desolve() so that it uses Maxima gensyms instead of Lisp gensyms for the dummy variable for laplace() and ilt().
What this means is that the variable names will still start with g, but they are now regular Maxima symbols (so they won't have the question mark).
As I understand it, this fixes the originally reported bug.
I've reported this potential fix downstream to http://trac.sagemath.org/ticket/9825
Closing this report now since there has been no further response. Feel free to re-open if needed.