On some tests (e.g. audio-encoding) the first run really suffers from disk activity and counting it into average seems a bit odd to me.
With the latest commit today, now if a test maintainer adds <IgnoreFirstRun>TRUE</IgnoreFirstRun> to their test profile, it will automatically ignore the first result and will not use that in calculating the average.
I don't know if it is possible to detect the driver settings of nvidia-glx or fglrx, but it would be nice to have AA and AF settings displayed in the results.
New feature... Now if you're using a NVIDIA graphics card, and you run a graphics test from either a suite or individually, if you over-rode the graphics setting from the nvidia-settings utility, the overrode Antialiasing and Anisotropic Filtering levels will be reported under the test notes. This is now in git.
PTS now also can monitor the die temperature, but I haven't decided how/where to use that yet.
On the GPU temperature, maybe it would be enough to have another optional test note. It would be very useful to compare different graphics cards (think of cards with the same GPU and clock rates, but different cooling solutions). Minimum and maximum value would be adequate IMHO, I don't think average calculation or visualization in a graph would make much sense.
On the other hand, this could also be optional for other sensor readings (CPU temperature, mainboard temperature, fan speeds...). Having an overall command line option for reading availiable sensor Information would be great, but I have no idea how to present the results in a sophisticated way within every single graph or test note.
There is another thing that came to my mind which I am a bit concerned about. PTS generates a lot of additional traffic on external sites (think of all the Nexuiz downloads). Wouldn't it be fair to have those dependencies bundled in an archive and distributed throug bittorrent or something similar? Maybe an included bittorrent downloader would help to reduce traffic impact and be more gentleman like.