Originally posted by Hound Tooth To be clear, I don't think this necessarily has anything to do with the camera's writing speed. I think what we're seeing is the image processing speed limit of the camera.
That can't be true, since the camera can process images (at 7fps) faster than it can write out - this is true for most cameras, where the camera can achieve a higher write speed in JPEG only compared to RAW, so the write speed is constrained by how fast the camera can write out the raw buffer.
In any case, the variance in write speeds between firmware revisions show that the current firmware is not at optimal writing speed (speculation about increased stability aside).
The actual write transfer speed is at 25 MB/s, which is driven off the HS clock rate at 50 MHz, so an effective write speed slower than 25 MB/s is a combination of card latency and wait time between blocks.
Since the K-5 runs a Fujitsu real time operating system, the slowdown in Firmware 1.03 could potentially be due to code changes that increase the wait time for some operations. As you say, potentially some of the code changes could be for increased stability, but equally it could be unintentional.
It may even be as simple as debug code left behind that could potentially be removed in the next firmware iteration. It depends on how serious Pentax engineers are about speed optimization - I suspect trying to fix the FF issue has probably taken up all their attention in this release.
Here's an idea for Pentax: release parts of the firmware into open source, and allow the wider community to hack it. Key routines that Pentax wants to keep away from their competitors can be in binary form only (encrypted if it's really critical).