One word of warning to those experiencing instability: if you are triggering writes to the internal OneNAND memory (the one where most of the system files live), you are better off with reflashing the whole fiasco image.

Because of the nature of the medium, write operations cause shuffling around of random pages, to even out the wearing.

So, if your system memory gets corrupted, chances are that the data being shuffled around will be corrupted as well in the process. Meaning that this might have side effects on _your_ data as well.

Imagine if a library is corrupted and this library is used to manage your phone directory ...

Sure it's unlikely, but it can happen.

About overclocking in general: even if you are one of the lucky ones with a silicon that can withstand the clock rate, the code will be executed in timing conditions diferent from what it has been stress tested with.

So you are likely to expose timing bugs that so far have simply been hidden due to specific timing making certain locks redundant. Nothing dramatic in this case - it will not kill the HW - but it can still corrupt your data.

FAT partitions (like the one containing data exposed when conencting to PC over USB) and all the memory cards in general (unless you have formatted it by yourself) are relatively easy to get in corrupted state.