Honestly, i wouldn't even know where to start. Changing the spindle motor may up it from 7,200 to 15,000+, but if the head controller can't keep up, you'll just get streams of garbled data. The spindle is the most obvious thing to change though, since manufacturers do this all the time, they just settle on 7200 as it's a good balance between power and lifespan. I know WD have the Raptor drives, but they are seriously expensive, and it's cheaper to get an SSD for much greater performance, albeit limited capacity.
What people have been calling out for over the years is a dual or tri-head configuration. Instead of one head per platter, you put another on the opposite side, one reads, the other writes, or both set to read, work on different parts of the platter to prevent I/O corruption, etc. But that just increases the number of files that can accessed at a time, it won't have much affect on latency - pretty much the biggest problem for IO performance on a hard drive.
I think reliability goes out the window when it comes to OC'ing a hard drive - and the file system will also affect performance too due to redundancy and checksums etc. I think that if it could be done reliably - the hard drive manufacturers would have done it already... maybe...
