Putting aside the problem of having the logic work differently for different countries and different size leagues and the problem of the massive upheaval it would take to initially set this up, a more fundamental problem is that not only does it create more levels that teams must promote through to reach the top, but it also dramatically effects the promotion-relegation numbers.
The current paradigm says that four teams will be relegated from I, and the four champions in II will promote. Although there's still the problem of teams tanking most of the season and then loading up for the relegation series, there's still inherent risk in finishing below fifth place, and the top league will have four new managers each year. In your proposal, there are two leagues below, so the two champions promote and that means only two teams must be relegated. This would likely lead to even less turnover at the top and slower progress up the ladder, as the possibility of falling out of your league is pretty much only something faced as the very worst team in your conference or by intentional decision. Of course, you could also promote the runners up or whatever else, but I like the "win your league to promote" concept. I also don't like opening up 8th place to a relegation playoff, which would be the other alternative - at that point the idea of tanking all season and buying your place another season is even easier, and I also think the idea of finishing last in your conference and not demoting is the wrong direction to go.
I think there are real problems with small nations, but the solution is more likely to be found in how new teams are apportioned into the lowest leagues. That's a side effect of stagnating userbases - I don't know if it's dropped enough now to no longer be the case, but in Hattrick the USA had the exact same problem, where teams would go 14-0 in VI, promote to V and get clobbered pretty much every game, and yo-yo back and forth for RL years. Maybe breaking up the leagues every year and making sure that teams are distributed across all the series at the lowest level rather than all clustered in III.1 would help? (I don't claim to know how BB does this, and they may very well do so differently.)