For the last question [edit: this part of the question has now been removed..], the relation of constructible sheaves with the Fukaya category is the subject of the paper Microlocal branes are constructible sheaves by David Nadler, and its predecessor Constructible Sheaves and the Fukaya Category by Nadler and Eric Zaslow. As for D-modules in general, there are proposals in the physics dictionary, starting in work of Anton Kapustin (see e.g. here) and perhaps best summarized in the seminal paper of Kapustin-Witten. One has to be careful though what exactly you mean: to get nonholonomic D-modules, the idea is to look at coisotropic but non-Lagrangian A-branes, ie not the Fukaya category but an enlarged version that has yet to be defined mathematically. Also even in the Lagrangian case there are two really different things one can mean by Fukaya category of the cotangent bundle, depending on how you treat behavior at infinity. Nadler uses a refined version that corresponds precisely to constructible sheaves. The one that arises in the physics, and in most of the math literature (eg work of Abouzaid) is the "wrapped Fukaya category" -- this corresponds more closely to modules over infinite order differential operators - in particular delta functions at distinct points on the base become identified (by exponentiating translation)! so this is quite far from what you might want in say representation theory.
As to your other question, I'm not sure I understand the issue -- D-modules form a perfectly nice stack, i.e. they glue together, independently of holonomicity (one doesn't need Riemann-Hilbert to see this, it's immediate from the definition as modules over a sheaf of algebras - in fact from my, perhaps naive, point of view it's easier to see perverse sheaves form a stack by thinking of them as D-modules, but that's certainly not necessary either). So the gluing formalism you ask for is just sheaf theory if I understand correctly: D-modules on a cover plus gluing data define a D-module on the total space. Same holds on the (dg) derived category level. This is in fact how you even define what a D-module on a stack is (in terms of smooth covers), as is discussed at great length in the last "chapter" of the text by Beilinson-Drinfeld on Quantization of Hitchin's Hamiltonians.
Now for the question of describing simple modules, that's a seriously tricky issue that I know nothing about -- I believe Toby Stafford was the first to show that there even are simple but nonholonomic modules over rings of differential operators, and I would look at his papers for insight.