Here are some thoughts on all of this.
There are, it seems to me, basically three channels of energy in the 1.9 theme system evolution at this point.
First, there is @martec's very respectable view that time is of the essence, and that storing the new Twig templates in the database rather than the filesystem, and leveraging existing eval-based code but making minor changes for Twig, would lead to minimal changes and a faster release date for MyBB 1.9. As I wrote to @martec privately some time back (I hope you got that message, @martec!), this is a correct view, however, since (as best I can tell) the majority of the long-term community prefers that themes (and especially Twig templates) are ultimately stored in the filesystem rather than in the database, this would require us to take one approach (DB-based) and then later switch to the other (filesystem-based), which, in my view, would both confuse users as well as waste (scarce) developer resources - better is to implement the ultimate approach first and for all.
Second, there is the approach strictly outlined in the document, linked to above by @Devilshakerz, which he (@Devilshakerz) has, with input from others, very carefully put together, and which @martec seems to agree has some very powerful ideas in it.
Third, there is the approach I'm taking which is deeply based in that document, except that it prefers a "Processing" approach over the preference in the document itself: the "Cold Duplicate" approach. My "Processing" approach ensures that the MyBB system is aware of all installations and updates of plugins and themes, so that it can archive old versions of the plugin or theme's themelets before installing the new version, as well as update any metadata as necessary. It provides dual themelet subdirectories:
devdist for development and distribution, as well as
current, for currently-installed version. The "Cold Duplicate" approach of the document in contrast relies on filesystem monitoring to synchronise the current version of a themelet with its archived version. I have a rationale for preferring the "Processing" approach, which I'm happy to share, but suffice it to say that I have made a *start* on implementing it (otherwise respecting the document which @Devilshakerz has put together, which I otherwise see no need to diverge from). That start is in this branch on my GitHub fork of MyBB:
https://github.com/lairdshaw/mybb/commit...eme-system
The fact that there are three channels of energy on this topic is both relieving (the dream of MyBB 1.9 is alive and well!) as well as worrisome (we have limited development resources; splitting them in three is suboptimal).
In any case, that's how I see things right now. Thoughts, comments, questions, and criticism welcome.