Here's a new pic.
I've marked the pins on both U3 and U4. For U3 I used it's data sheet and for U4 I used this
http://www.ti.com/lit/ds/symlink/lmr12010.pdf which seems close enough by the looks of it.
Dashed lines are connected through vias.
U4 pin5 to the ground is 200K - it can be observed that it has those 2x R104 on the way to gnd.
And those R104 are not shorted, the old pic was a bit misleading.
I repeat that C28 is not shorted as well.
TVS1 diode is fine - albeit the new pic shows 8v on both sides, in reality there was a voltage drop which I omitted.
I don't think driver chip U3 is a suspect yet. It gets supply of 25V and produces two nearly identical outputs of ~8v. It's vref is empty (this is the 3.3V lane) and it's inputs are also empty (because the main cpu is not powered, because the 3.3V lane is dead).
To be fair I fail to understand the logic of the board - how would U4 ever get any Vin unless there are some other connected points which I don't seem to be finding anywhere else on the board.
Q10 as soon as the USB 5V is connected, slashes the U3pin5 resistance-to-gnd in half - see green doted arrow in pic. Q9 - I fail to see the point of it at all - ground on the gate (when Q10 switches on), ground on the source too.
What's the use? What do I not know? Oh, ok, it seems to disable U4 (by supplying gnd to EN), when the USB is not connected, so the 3.3V line stays dark.
Anyway it seems there really has to be another place from where the purple/U4 supply line gets voltage. Will try to find it again.
Any hints where to look at?