Disclaimer/warning: long post:
Sorry to jump in here, I'm a new sysnative user, but, I hold formal degrees in both computer hardware and software and have been building systems and servers from the ground up for more than 30 years. That being said, I have seen nearly every type of "strange" computer behavior you can shake a stick at. I found your post/thread when searching for something else online.
Is this system under warranty? If not, based on everything in the post/thread thus far, if this were MY system I would be taking it apart (non-literally and literally both) to perform a *thorough* hardware reliability analysis, and would be dividing the system into functional blocks for reliability testing; power supply, motherboard, RAM, storage (drives), the CPU itself, cooling considerations (are all fans working at all times). Sometimes you can do this conveniently if you purchase/find a used *identical* system to logically swap/interchange and test components, then when all is said and done, if you've replaced any defective components, you also have the benefit of two identical working systems, i.e. a redundant computing set-up (if one system goes down, simply move to a different chair and continue working).
Disclaimer: everything I'm eluding to in this post may branch-off to something more complex and detailed which would require *your* time googling/reading/etc. That being said, you can use various offline testing methods and boot the system from a diagnostic CD such as the UBCD4WIN CD/project, which is bootable XP CD, and ISO image that you download, then burn the CD using a friend's *known good* PC and therefore use the CD to boot the suspect system and perform various hardware tests (first) on the suspect computer.
- A noisy power supply can inject electrical noise into the controlling circuitry on a MB.
- A temperature-sensitive chip on a MB can cause intermittent data bus/transfer errors.
- A temperature-sensitive or faulty CPU can "appear" to work but cause data errors.
- A defective chip on a Hard Drive's controller PCB can cause data errors.
- The storage device itself might be intermittent.
- A CPU fan or case fan could be quitting silently or intermittently.
It really does take quite an effort to narrow-down, weed-out, etc. a faulty hardware component, if it IS hardware ... a process-of-elimination in other words, but when you're done, you'll have increased the reliability of a system by a factor of 1000+. Off-the-shelf "entertainment class" PCs are not designed, built, nor tested for the kind of reliability that consumers really need these days, people nowadays are using workstations for serious financial computing or a home business for example. Many people would say they DO have a reliable, working PC, BUT it could work for 10 more years, or crash tomorrow. When you get or have a reliable PC, make sure to perform regular DRIVE IMAGE backups.
My own approach to strange-behavior systems, is firstly prevention, then hardware reliability testing/validation, then operating system validation, then driver checking/research, application compatibility issues, i.e. keeping a system minimal, then building it up slowly and watching for the reliability "crash", if any. If it were me, I would be rebuilding the system's software config. from the ground-up, carefully, after running extended tests on all hardware components.
This may seem like a big "dump" of information but I'm trying to point you in the right direction, for the long-run (fyi - I type fast). It would of course require your time and research. Again, sorry for "jumping in" please try and see the value of my post, all of which is challenging to communicate in just one forum post (or this post might also someday help someone else who is experiencing rare and extremely strange system behavior).