I agree. The only reason I am using 2.x is that they added some additional autopilot presets (which are stored in the registry). That allows me to use the CLI with specific models (using the workaround to apply the saved reg files before starting the CLI exe) - less than ideal, but better than nothing… But I still can’t force “always sharpen” and auto-detect never works…
Being a senior software engineer myself with a few decades of experience I have to say that that doesn’t have to be the norm. You can incorporate functional and integration tests, that ensure the functionality. Of course you can’t predict every eventuality (new picture input), but if you continually cover your bases, chances are very good that new versions still work. And the CLI is an excellent base, which Topaz Labs could use to build tests. I suggested something in this idea:
But I can only suggest things 
