FYI, the slim models without NEC/Tokins have aluminum polymer caps with a quantifiable service life as well. IIRC they are 2000hr @105c rated, twice that of tokins. But it's not magic, they'll degrade eventually too. It's just a matter of time.
The 40nm RSX can usually clock higher, so that's a nice bonus, but there are other reliability considerations for 25xx and onward models. Namely, the wifi/blutooth module and a step-down converter that powers it have a tendency to fail. 20xx and 21xx models appear to immune from that particular issue.
The ripple is primarily generate by the switching frequency of the buck converters. which is between 500KHz and 1Mhz. That doesn't change much and the bulk filtering caps (NEC/TOKINs) are chosen to smooth out the gaps in the pulses (2-phase for RSX). The filter is attenuated (combined frequency response of the bulk filtering caps and MLCCs is calibrated) to reduce switch mode noise and harmonic frequencies of it. The amplitude of the ripple is what we're concerned with. Because if it trips the over (OVP) or under voltage drop-out (UVDO) protection you get a YLOD with SYSCON error A0801002.
The amplitude, or how much the voltage swings peak to peak (think of it as a big wave) is affected by load. So the higher the load, the higher the waves (more ripple). When you overclock you increase the ripple that much more. From our testing in terms of FPS, up to 33% more. But I still need to measure the ripple on an oscilloscope to quantify how much more voltage ripple it generates, as it's likely is not a proportional relationship to performance. Just cuz FPS increase 33% doesn't mean ripple does.
Now, the operational frequency of the core/VRAM might induce noise on the line that exceeds the bypass tuning of the filter and cause instability. If so, then yes a new filter would need to be devised to address the high frequency noise component.
I need to revisit this myself as I have a console that does the same thing when the power is unplugged from the wall, even though the clock battery has been replaced. I don't think it has to do with the polarity being backwards, but sometime the connector pulls out and I forget which way it goes back in without looking at a donor board. You won't have that issue on a slim with the button cell socket.
