My point, which I guess I did not make clearly enough, is that container systems don't necessarily exist or remain supported over the ten-year period being discussed. The idea of ironing over long-term compatibility issues using a container environment seems like a great one! (For the record, .tgz -- the "standard" format for scientific code releases in 2010, does not solve these problems at all.)

But the "implementation detail" of which container format you use, and whether it will still be supported in 10 years, is not an implementation detail at all -- since this will determine whether containerization actually solves the problem of helping your code run a decade later. This gets worse as the number, complexity and of container formats expands.

Of course if what you mean is that researchers should provide perpetual maintenance for their older code packages, moving them from one obsolete platform to a more recent one, then you're making a totally different and very expensive suggestion.