The new issue #35541, inspired by this `ask.sagemath.org question reports a non-insignificant discrepancy between Maxima’s behaiour of its solver and what our interface reports, and may be worth a look by someone more expert than me…
As far as I can tell, this is a new problem...