I don't guess there's any real reason it can't work just fine. You, of course, have the normal loss of information anytime you redistribute from one protocol into another, but there's an argument to be made that you get that when you use network statements as well, so that's kind of a wash.
The main benefit that I see in this is that redistribution would give the setup the ability to react more dynamically to what happens in their network than the more static network statements. The downside would be that that dynamicism (is that a word?) would be harder to keep in control and might end up doing things that you don't want it to do. Of course you can put filters and route-maps and the like to constrain the redistribution...but then that's kinda like putting the network statements in.
Not that I would advocate being deceptive in what you're delivering to a customer, but if its fully managed CEs...would they even know the difference?
Assuming you're only talking about a handful of /24's like your diagram would suggest, I just don't think it makes all that much difference one way or the other. If you set up the redistribution to only allow the /24s, or set up network statements to advertise /24s...ultimately it ends up pretty much the say, right?
What may be the even bigger question that I would have (tongue-in-cheek), is...who uses RIPv2 in this day and age?!