Thank you, Chaoticone, these informations are most welcome.
Of course I can fix the lua script, but relating to #4001 and #4003, in my opinion a save and restore of a variable shall not change the value of that variable. In this case I presume the bug is located in the mach kernel itself. So "mc.mcCntlSetPoundVar(inst, 4003, CurAbsMode *10)" would fix the symptom and solves my problem with the build I use, but doesn't fix the bug itself.
Var #4014 is a different kettle of fish. With an active G54 #4014 in build 2914 is "1" and in build 3208 #4014 is equal to "54". So it looks like a different interpretation of the value. If this change is permanent the lua script has to be adapted. For this it is very usefully to know something about the meaning of the #vars. Do you, or someone else, know whether and where the #vars are documented?
Steve