I just looked at a 41C schematic I downloaded from somewhere (I can't remember where!) that has been mentioned here before - the chip labeled "CPU" "U10" has two pins labeled "OSC1" and "OSC2" and they have an LC tank (an inductor and a capacitor in parallel) connected to them. 82 microhenries and 150 picofarads which would have a resonant frequency of 1.44 MHz. If you reduce the capacitor to 1/4 of the original value, you double the frequency. Actually it would be a little less than double - the actual capacitance in the circuit would be 150 pf + the stray capacitance including the capacitance of the chip. This extra amount could be in the 10s of pf, so would be considerable compared to 39 pf (the closest standard value to 37.5). Rounding down to the next lower value, 33 pf, might offset the stray capacitance. The stray capactance would also affect the original frequency I calculated - I would need to add the stray capacitance to 150 pf. If it happened to be 15 pf, the resonant frequency would be 1.37 MHz. Changing the 150 pf to 37.5 (pretending you could find one) and including 15 pf stray capacitance would give 2.43 MHz. ( f=1/(2*pi*sqrt(L*C)) )
The schematic also shows two more capacitors connected to the "OSC2" pin. One goes to ground, the other goes to Vcc (the positive logic supply) and there is text saying "assemble either C10a or C10b for adjustment". I've learned today, looking at the HP97 service manual, that HP likes to use selected components to adjust things like the speeds of the printer and card reader motors.
The schematic doesn't say explicitly what version of the 41 it is and I'm not knowledgable enough to tell by looking at it. The label block has the pathname "d:\hp41\sch_pcb\nut\nut-main.sch". The reason I wonder about the version is that I think the CX ran faster than the C, although that might not be just because of a different CPU clock rate.
I stumbled on a website for gamers - they were overclocking Pentiums (and Athlons, I suppose) with Peltier device coolers on them, then cooling the Peltier with flowing water! I think just about any digital logic can be overclocked to some degree if you reduce the range of conditions in which it must operate - in this case, by controlling the temperature of the CPU. I seem to recall there was a 486 50 MHz that was not a DX - it came from Intel with a Peltier device without a fan, just a big heatsink, called "the Icecap", which kept the 486 die at 0 degrees C.