So the plot thickens. The v2016 files seem to act in much the same way, but this is with the TCv19.zip version.
I have isohidden paperspace open in Drawing2.tcw, 2 cubes showing in Hidden Line.
The viewport has "Override layers visibility" checked
Native Draw is GDI
In the Layer Palette, turn off layer "Left" - left cube turns to wireframe
Layer "Left" back on - left cube HL again - i.e. both cubes now HL
In this order, layer 0, Left, Drawing3|Layer_0 OFF - nothing displays (viewport is on Layer 0)
Layer 0 ON - Left cube = WF Right cube = blank
Left ON - Left cube = HL Right cube = blank
D3|L0 ON - Left cube = HL Right cube = WF
D3|L0, Left, 0 OFF then D3|L0, Left, 0 ON - Left cube = blank Right cube = HL
D3|L0, Left OFF - Left cube = HL Right cube = blank
Left, D3|L0 ON - Both cubes Hidden Line
So it is possible to get this to work, but so random as to not be practical. It may just be a v2016 bug that no-one came across, or found a work around for, but at least I hope this saves anyone else running into a dead end with a complex drawing. I guess the moral is to test any new technique on a simple model before committing.
How do other people manage an architectural job with multiple storeys and before and after versions of openings etc? I intend to carry on with xrefs but only in wireframe, and explode viewports from those external files for hidden line perspectives.
TC 2016 Pro Plat 64bit
Intel(R) Core(TM) i7-4700HQ CPU @ 2.40GHz,
4 Core(s), 8 Logical Processor(s) 12Gb ram
NVIDIA GeForce GTX 770M
Windows 10 Home