Okay, so here's what I have found (or rather not found) so far.
There are no traces coming from the fan connector visible on the upper side of the board whatsoever. There is also nothing that I probed around the connector that would measure 8V above ground (but I haven't probed everything -- only those spots where I didn't risk shorting things). So the fan supply circuitry must all be sitting on the back side of the board, and I am not willing to remove it, so we'll need to wait until someone else does it, or we can try to look for it in
Dave's teardown video.
Now, main rail is 15V, fan is running at 8V, so that is a 7V drop at what I estimate to be 80-90 mA, or about 600 mW, if a linear regulator was used, and that would mean a component of a substantial size to handle this heat (not a SOT23 one at least). There seem to be nothing like that around the fan connector, or, from what I see in Dave's video, on the back side, either. So, most likely, this 8V line comes from some switching converter, but if so, its output is filtered very well.
I probed the fan power with my old handheld scope, which is lacking in many respects, but it would allow to see noise up to a few MHz and a few tens of mV p-p. It didn't show it. All it displayed was the fan's interruption of power consumption that happens four times per revolution and seen as short voltage spikes on the fan power rail (AC coupled input is used here):
1/15.5ms * 60 ~= 3870 rpm, and this is exactly what I measured with a laser tachometer: the stock fan runs at ~3850 rpm. As I hold the rotating fan to slow it down, the spikes on the displayed waveform become spaced further apart.
Interestingly, the 120mm fan that I'm replacing the stock one with does the same thing: briefly break the circuit, or stop drawing current otherwise, exactly four times per revolution. I wonder what's that. Might be a result of how they are made electromechanically, or some stall detection and prevention circuitry. I have no clue.
The delay between the main rail and fan 8V rail being powered on when the soft power-on button is pressed is about 500-700 ms, the fan is powered after the main rail. It jumps to 8V from the start and sits there -- no 12V startup pulse or anything like that.