This kind of came up in another thread, I'll repost some of my thoughts. I really liked how it worked in 4th, so adapting those concepts to 5th is a good place to start.
First how it works in 4e:
The first option is feat multi-classing. Using this method you pick a class as normal and then take some feats to get basic class features from another class, with additional feats to swap your some of your powers( spells and attack options ) with the other class. Using this method a fighter multi-classing a wizard is no less a fighter for doing so. With full access to fighter abillities and defender features. This fighter doesn't step on the wizard's niche because it's still a full defender. This method is also very newbie friendly, as it's impossible to make a gimped build with it. This method depicts a fighter that learns some wizard stuff at a later date, or one who just dabbles in magic. This option actually creates a character very similar to a 5th edition multi-class that dips only a few levels into another class, just with no trap options at all.
The second option is hybrid multi-classing. In this method each class is given a half-class write up and you pick any two and combine them. This option is a little more complex and could result in less powerful charaters if you're not carefull, but is still less trap laden than 3rd and 5th's style. Each half-class only gives you the most basic class features. Requiring you to take feats to gain the more in depth features from a class. The hybrid character then takes his powers from either class, but must have at least one power from each class. Niche protection comes from the fact that the class role abillities are limited. A defender can only mark one enemy at a time, strikers can only add extra damage to striker powers, leaders can only heal half as often, and controllers need to pick controller powers to effect the battlefield. How well you perform as each half is determined by your power selection and the feats you spend on class features. This method depicts the true equal 1-1 multi-class that 3rd and 5th's 10-10 even split method does so poorly. It actually creates a character more close to the 1st and 2nd edition's multiclass options. Although I personally think 4th's create a more balanced and useful character on average.
Both of these methods are feat heavy, so single class characters also have more tricks and versatillity in their own niche.
I wouldn't expect anything like this for the One D&D main books. The developers could adapt these concepts for latter books however, increasing the multi-class options in the future.
As for adapting it:
I was doing a little thinking on how these concepts could be adapted. The hybrid concept wouldn't be too difficult, just make some half-class write ups and figure when and how to dole out various class features. The feat multi-classing wouldn't work in 5th. There are just not enough feats going around to support it. One D&D might increase the overall feats some, but if they want any compatibility with legacy products it can't be much. There is one place with the necessary design space though, subclasses. This would be very similar to the hybrid model, but more stripped down. You'd make subclass versions of every class and when it's time to pick a subclass you could pick the multi-class subclass version instead. Just like with the feat version in 4th you wouldn't have to delay or give up on class features from your starting class. The only real problem to smooth over is class combinations that are very M.A.D.. Two attribute dependancy isn't a problem, but three or more would be. So they'd have to come up with some rules in these half-and-subclass versions to account for it. None of this would be a quick or easy thing to implement. They'd have to devote some real time to playtesting and balancing for it, but I think it would be worth it to expand the multi-classing options.
