![Laughing :lol:](./images/smilies/icon_lol.gif)
Just wanted to mention that the "cosmic ray" hypothesis can sort of be ruled out, if you have a statistically respectable number of save corruption reports coming from MP games hosted in the cloud, where obviously ECC memory is much more common than on consumer rigs...
Probably less relevant, but definitely more fun to think about in my opinion: if, you wanted to avoid a Poperian science problem where one random counter-example is supposed to "disprove" the hypothesis (which doesn't really work since, maybe some random cloud box just didn't have ECC ram for some reason) with enough save corruption bug-report data-points, you could figure out how strongly you can reject the null hypothesis that: "MP games hosted in the cloud are equally prone to corrupted save events as MP games hosted in end-user environments" by employing a χ² test. But probably there are not enough data points so I guess I'm just showing off my l33t st4tz skillz and/or mentally masturbating you by saying so
![Smile :)](./images/smilies/icon_e_smile.gif)