Sigh ... these posts have become a bit of a mess.
There are so many different bits and pieces I want to illustrate and I've been trying to squeeze them in around normal work. Worse still, because I keep leaving them then coming back to them and re-running tests it's easy to lose track of where I was, despite using more or less the same test scripts each time (any new scripts tend to be sections of the main test script). I suspect my decision to only pull out the more interesting parts of the output has contributed to the difficulties too, but with around 18.5 thousand lines of output, I decided that was more or less essential.
It has got so bad that I noticed the other day that there were a couple of significant errors in the last post which are easy to miss when you're looking at detailed output and must be even less obvious if you're looking at it for the first time.
The fact no-one said much about these errors reinforces my argument with several bloggers that less people read and truly absorb the more technical stuff than they think. They just pick up the messages they need and take more on trust than you might imagine!
So what were the errors? Possibly more important, why did they appear? The mistakes are often as instructive as the successes.
This is the tail-end of the subpartition stats at the end of part 5