One achievable advantage of adding solid state storage is the dramatic improvement in performance of 'journalled' systems.
In journalled/transactional systems, such as databases, journalled filesystems, there are choke points in the performance where you must be assured that certain flags hit the disks before and in between other data chunks. This can involve I/O-dampening seeks and cache flushes for hardware that doesn't support I/O barriers (typical desktop level disks). With the addition of solid-state storage, you can write the transactional flags to the solid state device instantly without the concern over write-caches which will lose data on power-off.
Thus HFS+ with journalling could perform significantly better by taking advantage of a small amount of the flash space. Databases could also improve, although this is less typically a desktop type activity.
One thing I don't really understand is the proposal to spin up the hard drive less often. Typical machines these days have around a gigabyte of RAM, which can be used to buffer both reads and writes. How would adding 256 megs of flash help greatly?
One problem I see on Linux for example is that the journalled filesystem implementations write to the disk periodically, regardless of whether any I/O activity is occurring. Perhaps this could help with such an area.
-------
To those discussing executing out of flash, it's too slow. The read speeds from flash are so much slower than RAM, that no one is going to want to do this for any sort of significant amount of executable code. Additionally, since you cannot put varible segments of programs (heap, data, bss segments) in flash, there is no real prospect for "instant on" by adding significant flash to the system. However, as demonstrated by the article linked, the lack of seek-delay in solid state storage can accelerate seek-bounded I/O tasks, like booting.
