Glad to see that zigzagoffset did some trick for you.
For the cleaning up edges, you can use profiling with adequate offset, or the Deburring op (drop down from the Engraving op). Which allows you to selectively process edges, pretty nifty.
Regarding stability, I do have some stability issues, but I recommend using the actual setup-sheets as suggested in sliptonic's awesome video tutorials. They will help stabilize your workflow and save your data!
Russ's work will not help much, he mostly works on activating the 4th axis, and if you're unhappy now, maybe don't open this can of worms
. But hey, if you want it...(also, it's still on python 2.7 look at my version, that's due to OpenCamLib being currently mostly 2.7, python 3 version of OCL is coming up soon).
I think besides the stability issues, your chief struggle is with intuitiveness (or lack thereof) of the PathWB in terms of "how do achieve that?". In my opinion there is currently almost too many options. And the spiral is still not the right way to go by the way. unstable as heck
That being said, your model proved to me that it's achievable once someone is used to it. I've had just as many issues contorting my brain using another CAM (professional, expensive as all) package. Learning curve and all that...