As far as non-native support for drag and drop tab controls, I should like to think it sounds perfectly reasonable with the resources at hand; I'm pretty sure people are just comfortable with doing it the way it's always been done. I'm not saying it's easy, or that it's as fast (which would be the secondary concern).
http://blogs.sun.com/geertjan/entry/breathing_life_into_a_dead <- This blog shows some of the nicer drag and drop tab features of NetBeans IDE, which is, for the most part pure Java, which to me means it's doable in .NET, even if Java code has to be ported(all necessary resources should be open source), it's doable. The existing toolbar docking manager in .NET should be modifiable, then when Mono has a .NET 2.0 implementation completed if you simply override many of the existing methods to provide additional support instead of subclassing and windows APIs it would just be a better design decision and nicer reusable code.
If there is concern for catching focus changes and the like perhaps somebody should look into the feasibility of using WndProc manually...
Beyond all that, I honestly think SharpDevelop could take some very beneficial cues from NetBeans. It's obvious they're backed by very well organized people and they've had a lot of progressive ideas. I'm not saying do everything they do, the NetBeans Platform is not the same as the SharpDevelop Platform (nice name though, right? lol), but many ideas still seam feasible. The concept of distributing "Packs" sounds great, with separated packages for CF, Reporting, Enterprise Pack, etc. It could diversify down the line so that there was a XNA Development pack.
I'd also love to see better ASP.NET integration and testing support. Ideally you should be able to host a server using System.Web.Hosting; Visual studio seems to use an external executable (<Windows Directory>\Microsoft.NET\Framework\v2.0.<*>\Webdev.WebServer.exe) which I imagine it runs in a separate AppDomain and attaches to the various trace and debug lines. I think the inability to debug ASP.NET applications (or provide code completion to aspx files, and it's incomplete template definitions) is probably holding it back a lot. Especially with service oriented architecture making a lot of headway.
Just my thoughts on the matter;