After all that, I did finally get on the phone with someone (call at the right time and they pick right up, 0 rings! Call at the wrong time...). The rep and I went over a board I'm working on, which really exercises push and hug n push routing features since it's got fields of tightly spaced vias to route through, lots of obstructing components to push traces around, and so on. At first, he offered some useful but ultimately not problem-solving workarounds (never knew about Route>Gloss selected, for example). After we opened both AD17 and 18 on his system and tried to do identical tasks in each, he agreed with me that the routing engine seems to have regressed substantially in performance. For instance: with a trace entering a dense sea of locked vias, terminating at one somewhere in the middle, push that trace around. In AD17, you can drag any segment of the trace, as you'd expect, and it'll find its way around the intervening vias as best it can, until you've created a 90* corner or hit some other impassable obstacle. Occasionally, it'll stop prematurely if there's no easy-to-find solution, but this is generally rare. For example, the trace snakes around a bit then exits the field out the bottom in a vertical segment. If I drag that segment left and right, it'll find some path or another to let me do that, preserving the angle of the dragged trace. In AD18, when we tried this same task, we weren't able to displace the trace left and right arbitrarily. Rather, there were only two or three discrete right-left placements of the track that the routing engine could successfully solve an escape path for. Huge regression in capability, and makes certain designs borderline unroutable when the layout demands you place a lot of traces first, then push them around to make space for others.