I'd like to underline one important aspect: If the software update is "continuous", there should be an option to postpone updates until they have been tested and tried for an extended period, and also skip an update or updates until later time.
I know for a fact that many larger companies (10+ employees) using eg. Archicad often skip a version or two, since there is some known "dangerous" bug in the newer version, or simply because they know from experience that updating right away might be detrimental to their end product (=flawless building designs).
Software reliability is of paramount importance in the architecture business. Even small two-person offices (like us) are responsible for buildings costing several million euros/dollars. It's a very different business from eg. graphic design if one thinks about the potential liabilities.
A small error in a worksheet etc. can easily cost thousands for the architect. I'll give you an example: In one project, GetObjectSpaceNum for object returned space numbers of a different layer than that specified in the worksheet criteria. To this day I don't know why. In some other projects it seems to work fine, with the exact same setup. Of course we check manually but one doesn't always notice everything. Possible costs due to error like this could be huge. Now I don't dare to use that function at all.
I personally love VW and it's versatility, however I'd like to suggest maybe a major version cycle of two years would also halve the number of initial bugs?
Now it seems every year very fancy new features (I mean, which architect needs that recent 3D twisting tool...really) are added with the cost of simple quality control. Currently I'm waiting until summer 2017 to update to version 2017 even though I'm in VSS, to get all service packs that are coming before starting with the new version.
Look at McNeel and their Rhinoceros: Versions can be several years apart, and the software is, as far as I can tell, extremely robust (however it's a different kind of software).
Perhaps another thing particular to the architecture business is that projects are of very long duration. It's not at all unusual to start a project this year and still be working on it five years later. At the very least the projects must tranfer seamlessly to the new versions, or one has to run five generations of VW's in parallel (confusing).
Still, I very much appreciate that Nemetschek takes user feedback into consideration so well. If you also up your quality control, you have a great product.