Motherboards are prone to bad capacitors, and these usually present with random hard lockups and crashes, wavy unpixellated lines on screen (integrated graphics) and failure to start up. This one was different...
The PC presented as "I think there's a virus, can you clean it up?" with negative response to queries about lockups and "big" system crashes. RAM testing in MemTest was OK; Scandisk C: found mismatched FAT, but passed other FATxx volumes D: to F: as OK, with no drama (errors, slowdowns or old bad blocks) on surface scans of these volumes.
I went back and approached C: via Norton DiskEdit. The FAT mismatch was a single cluster entry that chained into the following cluster in FAT2, but ended with an EOF in FAT1. I wrote a matching EOF in FAT2, and comparison of the two FATs then found them identical. I exited DiskEdit and ran Scandisk again, and saw this...
...which is the last thing you want to see when running a file system repair tool, even if it's safely interactive as Scandisk is, and ChkDsk /F most certainly is not. Note the progress bar extending off screen, and "274% done"?
So I pulled the HD and repeated the RAM check for 20 hours, with no errors...
...and did what I should have done first; eyeballed the motherboard capacitors:
Yep; two already frothing baked goo, and one bulging on the way.
The unusual point about this case is that although it presents as deranging in-memory processing (buggy behavior of Scandisk), it doesn't lock the system after hours of hot no-idle-loop RAM testing, and neither does it throw up any errors during this process. Simply passing an extended burn-in RAM test without lock-ups or resets and lacking a suggestive history does not exclude significant adverse effects from capacitor failures; that was the surprising take-home lesson from this case!
(C) Chris Quirke, all rights reserved, April 2005