And often there's nothing wrong with that.
We don't want 6000$ laptops, even if their error ratios are 0%. Especially because it doesn't cost much to make the software resistant to the errors, and even more especially, because the error rate is already insignificant compared to the error rate of the software.
And with the software it's the same thing, often the error rate is insignificant, not causing the uptime to fall significantly.
And yes, sometimes the easiest way to find the bug is by having a million people try it out on a million different machines.
I mean, FOSS lives on that; the only projects without a major beta out seem to be dead projects.
