Whoops, you posted the wrong page. The statistics page that Hacker News needs to read is the one about how the Central Limit Theorem doesn't apply to everything damn thing.
In particular for software, instead of gaussians we often have the sort of distribution of completion times where "if the expected completion time is T1, but empirical observation says it never actually got done in between (0,T1], the conditional expected completion time is T2, and T2>>T1": ie, the longer you work on something without success, the further away into the indefinite future the horizon of expected success recedes...
...with a sidenote about how no the CLT doesn't actually mean that if you take lots of samples of something the distribution of those samples is Gaussian.