Comment 4 for bug 1826607

To make my previous note clear: I am not interested in mutation testing the innards of the functions generated by the CLOS implementation, so your solution of rebinding *macroexpand-hook* to #'funcall would be just fine. You will have to be careful that there isn't user code being compiled along with that (the forms in eql specializers, the bodies of methods) because compiling that could visibly depend on the user's *macroexpand-hook*.