Part 1 ended with a suspected dead GPU after the previous owner got a "reflow" at a mall kiosk. Reflow sellers count on you never measuring anything. So we measured.
rail expected actual ripple 5V standby 5.00 V 4.98 V clean 3.3V standby 3.30 V 2.10 V UGLY, 340mV sawtooth 1.05V core — absent (downstream of the above)
A 3.3V standby rail sitting at 2.1V with sawtooth ripple is not a southbridge — it's the two output caps right after the regulator having given up. ESR meter: 11 Ω and 9 Ω. Should read under 0.5 Ω. The mall-kiosk "GPU reflow" fixed a bridge nobody had problem with and charged $80 for the privilege.
Four low-ESR caps later (6.3V rating, 680 µF, ~15 mm height budget under the palm rest — height matters, ask me how I know), the standby rail sits at 3.29 V dead flat. POST on the first try. Fan spins, drive lights, Windows XP boots and the little IBM logo looks stupidly smug.
Next: the hinge bezel is brittle and the LCD inverter buzzes on the left edge only when the lid is at 95 degrees, which is somehow the most annoying possible failure mode.