I really like the idea because it's a fair way to basically test drive a character. I also like it because if they're able to pull this off, people really won't have a reason to complain about DLC characters unless they're just anal and want the full cast (even the characters they won't use).
Test them out. If you like them, buy them. If not, free practice dummy.
If the site(s) and pictures, blog posts are any indication, getting all characters will only make the game that much better. They look to be doing their homework before they release this game (so we don't see a avalanche of patch updates). I haven't downloaded any games from PSN or XBL in a long time. I hope Skullgirls doesn't disappoint.
In this game's case the DLC is really legit. As a small indie studio, pouring this amount of work into each character severely limits the amount of characters you can create within the time frame of your budget. You can't just keep delaying a game for the sake of adding content, the money runs out. People need to eat. Selling the game as planned allows revenue to come in which can be reinvested to extend the project via DLC.
This isn't a case of a company trying to nickel and dime its customers, it's a matter of these people being extremely focused on their dream-vision for the product and not being able to meet it due to the reality of the work load. The only way to get closer to realizing that vision is to start getting money back before you go bankrupt.
As far as the DLC for this game goes, all I'm gonna say is this...
I'd much rather put down $15-$20 initially on an indie game with only 8 characters and then possibly end up eventually buying a good amount of DLC and spending close to the typical $60 retail price on hard copy games nowadays when it's all said and done than to spend $60 on what is basically the beta version of a game with 30+ characters and then end up shelling out another $40 for the actual full game and essentially spending $100+ on one game (that's not including the DLC, mind you) like Capcom's recent business model seems to be fond of.
The former method doesn't make it seem like I'm being cheated for one thing.