I guess AutoCAD code nowadays became so huge and complicated that I bet they may even have fear to touch anything in that 'legacy' code.
In last revisions they made some steps to clean the internal structure, like the clear distinction between 'core' and UI functions, but I fear that was just another good initiative that will never get to a clean conclusion.
I've been working with AutoCAD products since... let's say European version 2.11, so I have some experience from this side of the barricade... and I know there are very good men working on it, just think at what we have today, so huge steps have been done, but there is something that always bothered me, expecially with latest 4/5 versions: they have good start point in some direction, but they do not complete the process...
Some examples?
- .NET support had had huge improvements between releases 2006 and 2009, but since then very little has been done. Not every ObjectARX classes has been exposed yet, some very important like the Associative Framework;
- in 2012 we saw the new Model Documentation functions, but since then hasn't been improved, they remain something not very well integrated with the rest of the environment and they haven't been exposed as API;
- we have been introduced to acCoreConsole.exe in (2013?), but it's not yet officially supported and documentation is still the same since its introduction;
I do not mention MAC support because I'm not current... is still there?
What I say is that sometime AutoDesk should sit at a table and just look back and try to clean things up. I work in Europe and despite the huge steps we made for living in a ultra broad-band world, real daily job is still very far from cloud services, points cloud, Forge platforms of VR handling... everything really fun, but not today.
I understand the need of R&D for future needs, but here the suspect is R&D is getting all the resources available... and it's not just an ipotesi, since I clear remember an answer from ADN support where they apologize for delay in answering because they where busy with the Forge platform event... so? you pull men from a paid service and allocate them for a event of a still-to-come platform that probably I'll never use?
I can just make speculations about what's happening with the development teams... 
Today we need a faster and more stable AutoCAD, that may use some more of the processor power we have now;
we need an AutoCAD/DWG format less prone to errors and damages, I don't want to spend hours in Auditing and Recovering;
we need current command set to work well and well integrated;
we need a development platform with a clear roadmap ahead;
we need quick and clear support to emerging standards, like IFC file formats import AND export, I cannot afford to buy and learn Revit just for a low quality BIM model (I know you'd like I did!);
and so on, I bet our fellow colleagues would add dozen of point to this list...
By the way, where should we publish a list like this? I've seen the IDEAS section of the site, but... where is the AutoCAD section???
I don't expect an answer from AutoDesk, after all this is not even a question... but it'd be nice if they'd read it...
Sorry for the outburst! I feel better now 😉