The clue is that flashing red light. Figure out what that is, and you'll have something to work with rather than groping around blindly hoping to stumble across the solution.
Also, if the unit shuts down within seconds after successful POST, how are you able to run memtest on it?
Gateway E-4100 mini-tower running XP Pro SP3. P4 Prescott 3.4 HT, 2 GB RAM, HDD HD5728040PLAT20 & Maxtor 6L 160P0, both refurbished to new condition. HD5728040PLAT20 is a slightly larger clone of the original WD400 40GB drive and is the boot drive. Clone was successful and is running in the system now.
Also running two DVD/RW drives, one is a SATA primary and the other is an IDE secondary master.
Video card is a nvidia 6200 agp 8x new and psu is also new: Corsair TX650. New psu replaced original Gateway 250 watt unit in early March of this year.
Motherboard is an Board: Intel Corporation D865GLC AAC28705-409 and the system model is Gateway MIDWAY 4000888.
BIOS is original: BIOS: Intel Corp. BF86510A.15A.0066.P13.0404221042 04/22/2004
Both drives are reported as healthy by DiskCheckup 3.1, except the Maxtor is reporting spin-up time is around 10 seconds to start.
With the case, open, when the POST failure occurs, a red led light will flash at the bottom of the motherboard and stay flashing. No video screen with the Gateway BIOS screen will appear; screen remains blank.
Force-shut down in this condition will take about 3-4 seconds; waiting 30 seconds and cold boot will be successful as normal. Red led light does not flash, and goes out in 1-2 seconds. Case fan will run continuously when POST fails, but will run 3-4 seconds with successful POST.
When a successful POST occurs, system shut-down is immediate.
Troubleshooting has involved replacing the original nvidia 6200 agp 8x card, and also booting without the HDD's connected. Also have run memtest, reseated all connections and cards, no change. memtest showed no errors after 7 complete passes.
With the HDD's removed, 1 out of 8 cold starts resulted in a POST failure. With the Maxtor removed, there was no change in POST failure rate. With only the boot HDD running, no change.
Still got POST failure when running HDD's, force-shut down, cold-boot then successful and system runs normally. Applications snap, no degradation of system performance noted.
Have checked capacitors, all look to be in good condition.
Reason the original psu was replaced was because the 12 volt rail was varying from 11.14-11.97 under load, even tho the load was reduced; new is 12.14-12.17 volts with all peripherals connected. Old psu was running with only one HDD and one DVD drive, and the original 6200 card.
Where to purchase a pci post card? Would using this device help track the POST issue down?

Chowhound
Comic Vine
GameFAQs
GameSpot
Giant Bomb
TechRepublic