Hehe; how about adding a feature to mkmerge that lets the user enter the number of MBs that he's willing to sacrifice for that individual file he's going to mux, and depending on that value there's more or less recovery information stored? Flexible, is it not? Hey, just trying to be a nuisance... :p
i'm quite convinced it will come down to that sometimes. However i expect it not to be as simple, because there could be different safety mechanisms, like header repetition and recovery information in each cluster. So there would still be the question which ones to use to reach a certain amount of MB of reovery data. And some mechanisms like recovery info in clusters would take a fixed amount of space per cluster, or per frame or per whatever. Now you could for example give the options "use cluster level protection" and "use header repetition to fill up to X MB of recovery data". ("disclaimer":I did not follow the in-detail discussions on error recovery and this are just my thoughts without much prior knowledge :p)
Guys, please help me bugging alexnoe to look more into this. I was trying to convince him he si the right guy for that, but it seems i was not successful yet ;) ...