Mmm.. and there's also this note I placed in the kernel source code years ago:
Code:

---

        // Mmm.. bit of a race condition here:  we shouldn't be accessing
        // the flash memory directly (playing the animation), in case empeg_state
        // wants to write to it when accessed via hijack_init() from display_animation().
        // So what we do about it is.. nothing.  Probably not an issue anyway.

---

I think a long animation would greatly increase the chances of this situation arising and causing problems. But there's not enough info to know if that was the case here (or is there? I didn't see any sign of a "bad flash checksum" message in the logs).

-ml