I’m not understanding what you are trying to do by using ConvertToSingleSpan. When I tried it on your geometry the command line responded with “nothing done”
I don’t know much about ConvertToSingleSpan command but it looks like it does pretty much the same thing as ConvertToBeziers. Its kind surprising that they created a new command that does the same thing as an old command.
So I`m trying to do your & Pascal way to use ExtrudeCrvTapered with a tiny depth and then use extendSrf (which doesn’t work over joined polysurface).
So using v8 ExtrudeCrvTapered with splitAtTangents turned off is bad because I can extend 500 separate surfaces joined into polysurface (too much time).
So I`m left with using ExtrudeCrvTapered in v7 (single surface as output) or using ExtrudeCrvTapered in v8 with splitAtTangents turned off.
In both cases, after doing the extension I try to split it into single spans to see if its true G1 or I do DupEdge on the back edge to see if that edge is full G1 ( I assume if the back edge & front edge will are G1 then is a big chance that surface is also G1).
So to split that surface into single spans I have a few options:
Why should I use that single surface produced from ExtrudeCrvTapered after extension and pretend that is G1 if it’s not a real G1 in my opinion what is visible after using those 3 examples or after DupEdge on that surface where small curves are not connected with G1 in some points.
So my question is:
Is there any way to produce a tapered surface from G1 input and real G1 output where the real depth of it is about 30-40mm without recreating all blends or rematching almost all edges which in bigger projects could consume a few days? Next, after some corrections, it is needed to redo the same job for another few days. During those hours I questioned myself how could I do faster?
I could accept workarounds but which will give me G1 as a final and which will not force me to fix manually surface after surface on 500 surfaces.
OK so ConvertToSingleSpan is just another method you use which is equivalent to DivideAlongCreases’
I have been complaining about the same problems with ExtrudeTaper that you have explained so well for 25 years. There are some strategies that you can use to counter the defects of this command, but frankly I wouldn’t pin my hopes on the developers suddenly becoming competent and fixing it.
I didn’t mean to suggest that workarounds would do anything more than make the process a little less painful.
Rhino has unique CAD tools and IMO should polish them more. I know that is not popular from the marketing side of view but in the time of Blender, Inkscape, Gimp you can do good subd mesh, texture and render literally for free. There is also Freecad and Plasticity with better and better Parasolid tools IMO Rhino team should invest more time in polishing its real gold- unique modeling CAD commands into perfection and leave mesh modeling for other software. That is my point of view because for mesh modelling I use Modo and Blender and when Rhino CAD commands fail I can`t get help anywhere. If mesh modeling fails I can use Zbtush, Modo, Blender etc. I hope someone will notice that and v9 will be only for polishing existing commands into perfection. Performance to work with 3d scans also could be nicer 
The next time you need this functionality, I updated the script so that it can now align the ends of each pair of taper start and end curves. Some angle accuracy is sacrificed to do that. If more accuracy is needed, insert knots in the input curves.
I will do a thing with tapered sides today or maybe tomorrow. I will check it with pleasure. Thank you so much for your effort.
I haven’t written it here but even before your update, I still preferred to use your script (spb_TaperFromPathOnVerticalFrames.py) instead of the original Rhino ExtrudeCrvTapered command. It`s much better and much more ergonomic in use. Output geometry is also cleaner even If I had to correct many blends it was much easier to correct.