Yes, we want address this problem with an extra tool. Our system is fast, but has atm no fallback to the standard language. So we want to provide a tool which automatically fills the empty descriptions with the main description and a tool, which shows which descriptions in the languages are the same, so that you can fast find non translated products.
Atm you must use the multilanguage system consequent to get the right consistence. As far I can see, we gained speed compared to other solutions, but loose the fail prove status regarding the language.
It is almost always in the informatic. You can force consistence with a lot of tests and workarounds and this slows, on the other side you can write faster solutions, when you just assume that the input is correct.
We always try to put the tests into the storing commands, because the browsing, the get functions must be fast. So you can have a slow and safe design or a fast nasty one :-). We decided for the multilanguage to choose a fast, but nasty one.