I can see why people are wary of this sort of development model, its really not been used with games so far. Its also had quite a few failures on Kickstarter that didn't deliver so there is a reason to be concerned.
The software industry changed quite a lot in the last decade with web software. Websites don't have any need for version numbers and most companies running websites moved to continuous delivery. That typically means they update changes to the site on a very regular basis, as much as every hour or faster and the change never comes with any update documentation or big version number change. New features just sort of appear. Continuous delivery matches the development really works at its core, we developers make small pieces and include them into the product on a daily basis and in the end the accumulation of these changes makes a final product. But on websites we don't need to make the final product every 12 months, we can release it whenever we like.
A number of companies have pushed the concept to client software as well. Chrome is a classic example as no one ever knows what version they on, the software auto updates in the background and usually you don't even know its happened, you just start it the next day and its changed. Occasionally you get a prompt to restart and that is about it. Well games can be done that way as well, and a number of games have been getting developed in that way in the last few years (project cars, Minecraft). They aren't super agile, like daily or hourly but weekly or biweekly is quite common. The main point is they don't wait until they have the game finished, they release what they have and smaller and bigger features roll out when ready.
What is interesting is the only companies able to do this are the ones funded by the customers and not by the publishers. Big publishers see this early access with continuous feedback from the customers as a risk to the schedule so they aren't interested. Yet its interesting that most of the companies doing crowd funding are trying this model, but not all. The guys making the syndicate remake for example have various insights into how they are getting certain things done but its reasonably irregular. When compared to the constant feedback cycle of star citizen its positively archaic.
Star citizen is still early, they are not at the Minimal viable product stage yet. What that means is there isn't a game or thing to play. They developed the hangar module, that worked and since they have been updating it continuously. Once the dogfight module is out they will do the same thing, it will start its cycle of feedback. You need a minimal product to be able to get feedback otherwise the discussions are never ending. Its not 100% in the spirit of agile development but its about as good as a game like Star citizen can get. They fill the gap with questions, comments, design discussion videos etc that show you their thinking so you can feedback on what they said.
None of this means they will deliver a game, but the contents information and decisions does build confidence that they know what they are doing. Agile requires confidence in the team and its built by continuous delivery of product, and in this case by talking about the building of their MVP.
I don't like the way the funding has been partially used to produce an early Pay 2 win scheme, I am still worried the very expensive ships will give others an early advantage. But I can't fault the way they are going about development so far, its not dissimilar to what I recommend all my clients do.