OK, the only major question I have at this point is where is 'infiltrator' really coming into this. You've got themes driving most non-combat skill/etc functions, which is OK, I think 5e will NEED to do more of this kind of blocking together of options and creating more smaller lists of things you can access in different combinations. I'm not sure themes need to be ALL about non-combat or restricted to that side of the fence, but that's really a different discussion.
I'm not sure that "non-combat" roles need to be strictly limited to non-combat either. (Certainly, there are stealth uses of combat that make sense. Most of the other skills have some combat-applicable utility power.)
However, I think the non-combat roles need to start with non-combat. 4e made a significant advancement over 3x by removing arcane magic as the ultimate tool to bypass any non-combat challenge. However, I don't think 4e really put anything back in its place. Skills are important, but dominate or are dominated by rituals and utility powers in something of a hodge podge. I'm not particularly set on any particular design, but I'd like the 5e designers to look at the major types of non-combat challenges, figure out the collections of abilities well suited to solve those problems and let the players pick from appropriate packages.
Anyway, I just haven't really found a spot where I need to attach some actual mechanics to 'infiltrator'. Nor does it seem like it does more mechanically than act as a flag.
Well, yes, it's a flag in much the same way that Defender and Striker are flags. Striker doesn't have any specific mechanic. It's just a marker to that says this class has a special damage enhancing ability. The role indicates that the classes (or themes) that are of this role will be especially good at the actions associated with the role (buffing/healing, defending, single-target-killing or AoE/conditions) or (infiltrating, persuading, exploring/traveling, or information gathering).
That's OK, but I'm leery of lists. Once you start making lists you are staking out some ground and saying "this list of things are the relevant things, everything else is right out". With combat roles I think that is OK because the 4 roles really DO cover all the conceptual functions of combat (and really if you study military tactics map fairly well onto the functions and concepts that say the Army will tell you are relevant). Any list of non-combat roles I would think would necessarily be open-ended, and as I say, I'm leery of open-ended lists where the list claims to categorize everyone. An open list of say races is OK, nobody ever has trouble fitting into that list. A list of non-combat roles OTOH might not fit everyone, and then you run into where players have issues or the system starts to feel restrictive or artificial.
I don't think the 4 roles are actually a complete list of combat abilities. I just think they're good enough. One could imagine a high-defense character class that creates large zones around them that damage and imposes conditions on both enemies and friends, but slightly more on enemies and even more when they are close. (Imagine some partially controlled elemental magic.) Such a character would be a lot like a controller, but they would also operate much like a defender because they need to go up to enemies and lock then down. There are a gazillion variety of possible classes, so I'm not really worried that 4e doesn't support every possible PC concept, but the existing roles is just one more imperfect framework that we've become used to.
But stepping back for a moment, I think the missed the major advantage to having non-combat roles, which is that it lets you design non-combat situations that everyone can particulate in. Sure, a face-character is going to dominate in pure persuasion gameplay, but you don't have to design the encounter that way. If there is a duke that needs to be persuaded, the GM can provide several important clues to help the persuasion that can be gained by information gathering, exploration or infiltrating, respectively. That means that each type of player can participate in the effort.
In effect, if a GM designs a situation where each role can contribute, and each role is guaranteed to be at least marginally capable in role-appropriate activities, then each PC is assured that there will be useful ways of contributing out-of-combat. That is very different from today's fighter who can easily discovery that Athletics, Endurance and Heal just aren't useful this module.
Of course, not every obstacle needs to have four approaches. (Not every lock or trap needs to be persuadable.) You could easily fill an adventure with a dozen obstacle, each of which is amenable to 2-3 approaches. That provides balance of opportunity in the aggregate, even if individual encounters/obstacles are dominated by a specific role.
-KS