On the mechanics side, I'm very much in favor of converting ground unit training into a system that works more like industrial production, but with minimum training times for specific units (similar to how the build queue works in Hearts of Iron). As it is now, using multiple GFTFs to train a large formation involves breaking down the formation template into individual sub-formations and fiddling with the construction costs until they are about equal, then training them separately and combining at the end. This is tedious and error-prone, and makes IMO only limited sense.
This has been a frequent request, and Steve has acknowledged the need but not yet developed a solution as far as we know.
I think a large part of the issue is that when Steve developed the ground forces system, he probably had in mind that battalions would be the base formation size (~2,500 to 5,000 tons) as in VB6 (you can see this in the rank themes as well). With this assumption, the ground forces training works without too much trouble as a 5,000-ton infantry battalion can be built in about 5 months at the base training speed tech level, and a similarly-sized armored battalion in 3-4 times that. Once you start developing the training techs these times go down and more expensive units (heavily armored or with added capabilities) can be trained in roughly a year to provide a natural progression of sorts.
The problem has been that the ratio of ground force leaders to number of formations with such small formations is not favorable, and so the equilibrium tends to settle closer to 15,000 to 25,000 tons regiments/brigades as the base size which take multiple years to train. Steve has noticed this, as NPRs tend to use ~15,000 tons as their base formation size, but a change for ground unit construction has not yet come. There is some hope that the doubling of naval and ground commander birth rates in 2.0 can help with this, maybe reducing the needed formation sizes to ~10,000 tons which is more manageable once you research a few levels of the training tech, but it's not going to be a real solution.