I think this Xkcd strip is missing an important point, which is that voting is exposed to a very broad attack vector: probably quite a large percentage of the p

I think this Xkcd strip is missing an important point, which is that voting is exposed to a very broad attack vector: probably quite a large percentage of the population would try to corrupt the results if that could go undetected. The risk that electronic voting is exposed to is corruption, not really accidental failure. The other examples in the strip are talking about accidental failure. So while it's true that software quality management is usually (very) bad compared to in engineering disciplines, this is largely because of missing incentives and arguably missing regulation (for worse in this view), which could be changed, but unlike in the engineering cases, it would be a different level of difficulty for regulation to solve the issue, since the devices in question can be exposed to bad actors on pretty much every level. Or with a corrupt government / dictatorship even the regulation agencies themselves could be compromised.