installations vs visits
There seems to have been some confusion between new OS installations and people accessing websites :
"we're told the stats were gathered from the logs of some 160m unique web surfers hitting 40,000 websites in the pollster's analytics network: each visitor's browser is expected to reveal some basic information about their computer"
I know one can be used as an approximation of the other but this will contribute to the monthly variation others have already noted. I expect another seasonal factor will be students going home and switching to their favourite fondleslab.
I also doubt that many kiosks access access the websites used in the survey , so don't think they will be showing up in the stats.