I'd argue the 3 most important gaps for SONAR in terms of getting more credible are:
1 - Performance. It shouldn't be the be all and end all but, like cars having great 0-60 figures, it matters. SONAR needs a decent audio/VST performance boost so it can compare favourably with others (which it doesn't in the unfortunately commonly quoted performance tests).
2 - Compatability. There are issues with VST compatability still - it's not bad, but there do appear to be more issues with SONAR than other hosts from reading various forums, and SONAR's methods seem to be criticised (wasn't it Acoustica who recently said that SONAR's thread handling was just wrong which caused problems with their plug-in?). This then reflects in plug-in companies testing with SONAR, which reflects on its popularity which reflects on companies testing with SONAR which... Same applies to control surfaces: Cakewalk should bite the bullet and write the surface dll'ls, not expect the controller manufacturer to do (I'm pretty sure that it's usually the DAW manufacturer who does this).
3 - Reliability. This is a tricky one as - apart from the rare issue I've seen, usually due to plug-in compatability and the Bakers erring on the side of caution with dealing with errant plug-ins - SONAR is no less stable than anything else (hell, I get Windows Explorer crashing more than SONAR!). But there's a perception of it being more unstable, crashing more easily and being more picky about hardware than other platforms. On shared forums, I don't see reports of DAW x stuttering and glitching with one or two audio tracks: I do see these reports about SONAR. This probably ties in with performance above to some extent.
All of these are technical issues that have a direct impact on word of mouth marketing and advertising (especially point 2). They're non-trivial to solve or they would have been, but would need a technical, marketing and charm (money?) offensive to address - eg. Pick a big plug-in manufacturer who doesn't currently support SONAR and work with them to get is supported, learn from them why SONAR is 'difficult', pay for their testing/technical expertise and re-write parts of the code accordingly.
Financial investment in plug-ins is probably greater than that in the DAW itself (all of which are ridiculously cheap for what they offer), so a professional (in terms of making a living from recording/mixing) will choose to follow the plug-ins rather than the DAW, so getting that relationship and compatability right is probably the first step.