I'm feeling a bit silly right now, haha! XD Why oh why didn't I think of doing this when I had a look earlier!

It's great to see that you've managed to figure out a workaround for this :)

As you said, from what i saw when I tried solving this earlier for Daniel, it does look like it gets quite a bit more involved as soon as you try to actually make the code handle the non-uniform case. From what I could tell, the only alternative to doing to would have been to fix the way that we apply the IK solver results back to the bone - i.e. how we calculate and apply the "ikmat" stuff - but figuring out how to do that looks like it could get pretty nasty. It also seemed that the IK solver itself seemed fine as it didn't seem to know (or be given) anything about the scaled nature of the bones (unless Brecht could confirm otherwise).

So, from my side, this will be fine :)

From what I can tell this change doesn't even affect what the IK solver does, it is already supposed to ignore such non-uniform scale. The IK solver itself only knows about rotations and translations.

If you change the non-uniform scale work around to only affect step 6 it still gives the same result, step 5 is not affected by it. It seems to affect how scale is inherited or something like that?

A problem with this patch is that it discards the scale on the X and Z axes of bones. In the example file they are all 1.0, and if you change them to something else you'll see it has no effect, while without an IK solver scale on those axes has an effect.