For me it's a bit the other way round. We also have build a deployment toolchain for structures/templates and have put a lot of work into a (still) clumsy process.
We found that editing them directly in the frontend is highly problematic. To resolve the issues we had, we created a extra tool here that allows me to sync templates between servers, allowing me to see differences between environments. It happened quite often that changes were done directly in production. Or only on the testsystem and left there to rot, because they were forgotten.
We still use that for hotfixes and to clean them up later. But now we enforce that all templates and structures have to be in git, all changes need to deployed. Sure, Liferay also versions them under the hood, but we never used that. It has some disadvantages, but if you have a large system, it is impossible to manage otherwise.
So, for me the new fragments toolkit is quite THE improvement. We did not really use it so far, we will start the first project in Mai. But with the toolkit we can simply have all code in git. That's the first great thing. And we can use scss and autoprefixing and ECMAScript instead of plain javascript all in one project, together.
I am mentioning this because we currently have two projects for 7.0. One for templates, one for javascript. And some changes often had to be made in the theme. That was the best solution from a technical perspective, but it didn't make sense from a module perspective.
We will probably struggle a bit to find the best practices, but in general we here expect things to improve.
Please note that we don't expect structures/templates to go away, but things will change. Now we have 50 templates. I guess, we will have 5 templates and 20 fragments in the future (I believe, a lot of the templates will become obsolete, since they can be build from smaller fragments).
