So, it can be done but politicians are preventing it. Market is bigger problem, though. Burroughs B5000 was first, high-security system nearly immune to code injection in 1961. They did good in market but IBM had better price/performance. That plus more feature improvements is all market will really buy. Something uglier, slower, a few less features, and cost double for the security? Most wont go for that. They usually dont do it even if the software is free and easy to use like a few in FOSS with strong security. So, suppliers are right to keep giving them the garbage they demand until they'll pay for something better.
Note: Companies and CompSci people continue developing higher, security stuff anyway. Just very few of them with commercial stuff costing big $$$ due to low volume/demand and academic stuff delivered in prototype form since FOSS people dont put time in them. Hard to see how it will get better on a large scale.