Well, personally, I must say that I know that the current implementation I have in LiClipse for StyledText doesn't have exactly the same behavior (which is something I'm already working on to improve), but apart from that, I must say I only got positive feedback when compared to the native one on dark themes.
Also, I may be wrong, but I believe users find theming MUCH more important than being 100% compatible with what the system offers, and is one of the main reasons people which use dark themes flock elsewhere in my view -- they end up choosing different IDEs and UI libraries because SWT doesn't do something which is a common-case for any modern UI library.
And in this situation, if users want, they can always fall back and use the native version if the themed version has some quirk... meaning: initially, I'm much more concerned in having an approach which will be accepted by SWT and having something working which will give users an improved experience on dark themes -- after that is agreed upon, it should be a matter of iterating and refining the implementation to improve on issues which are raised (and the native implementation should keep on working anyways, so, in the worse situation, users can keep on using it if they want).
So, in short, I know that emulating the behavior is hard, but it's something which I'm up to do, provided that SWT devs are Ok with the approach (one thing I don't want is doing all the work and not having it integrated because the implementation should be different from the start or because ideologically it won't be integrated anyways, which is why I'm bringing up the discussion).
Best Regards,
Fabio
p.s.: The current implementation StyledText scrollbar in LiClipse has been available for a full year without any critical issue -- although as I said, I know a few differences in behavior which I intend to improve to make it closer to what the system offers, even though users of the LiClipse dark theme seem satisfied with it.