Hey, Silvio; my pleasure!
Thank you for the kind acknowledge. This is gold... I must confess that when I saw your post, my first intent was to ask if you and your wife had a nice trip back to USA and about that babe; good news, indeed!
Just a question: are all of these 41 the same type? Full or half nut? I know that the 41C you sent me is a fullnut type, but the others share the same architecture?
Also, if you run a program the first time and it has GTO or XEQ to "short-" or "long'-"term LBL (2 or three bytes), it's gonna be a lot slower than the next executions because the 41's O.S. will compute the number of bytes necessary to "jump", forward or backwards, in the first GTO/XEB occurrence. This does not happen with XEQ"ROMfunction". And if you pack the calculator memory, all the process is repeated.
I have not felt significant difference from any of the HP's (C/CV/CX), but I guess the CX MAY (just a guess) be slower because of the internal circuitry. Also we must consider that the HP41CV has more memory space and more circuitry than the 41C. But based in your measurements, the 41CV is even slower than the 41CX.
The last argue: what modules were installed in the calculators, if any?
I have some knowledge about HW, and loosing clock cycles is a potential cause of crash in clock dependant devices, mostly earlier architectures, and this fact leads me to believe the only reason for the difference is an acceptable "range" (2 to 5%) in component values, causing a slightly different clock for each machine. You may measure two 41C's running time and find different values.
A few weeks ago my HP41CX scared me a lot, but I believe it's just a matter of failure in the batteries' contacts. I believe that sort of "bad behavior" will only be observed in an HP41CX or an HP41 with a time module.
BTW, would you help me finding an HP41CX, fullnut type? If you believe you can, please, e-mail me.
Grande abraço! (let's say it's a version of a closer "Cheers!")
Luiz C. Vieira