I hadn't checked vs 0.94. Yes, these are useful features that should be back-ported. +1 for that.
Can't blockSize be effectively set via -D option?
I'd also like to add a --sampleSize option, at least RandomRead, because it runs a looong time if you're running it against a table of any size. I have this feature in my own branch, which I've been using locally. I have it dumping the times taken by individual test threads so that they can be re-assembled for request latency analysis.