So, since I really was dragging my heels I thought I'd take a little detour. Instead of writing the UI as a layered frame system with an orthographic projection, manually manipulating vertex buffers to layer the elements, why not just do the UI as a 3D scene? I set up a second scene+camera+viewport, and configured the render path to clear only the depth buffer. This second scene is drawn after the main scene. UI elements are added to this scene as models and meshes. Text stuff is built as a custom geometry, character portraits can be the character model in idle animation, etc... UI activation and deactivation can send elements swooping in from off-screen then swooping right back out again. I think it could be kind of cool and fun, so I'm going to tinker with it the next couple of days. Hell, if it gets me out of having to mess with the UI layer code anymore, that alone would make it worthwhile.
Anyway, here is a current work in progress shot. I only have some test UI text, a test alert sent out to all entities to test the floating combat text, and a character portrait of GC standing in T pose (not animated, since I don't actually have the idle animation exported to Urho format yet.) If I'm not too busy at work tomorrow, I'll whip up some placeholder UI assets and start connecting those things up to the various control bits and see how I like it.