They seem to lead to the development of 'umami' flavours through proteolysis.
This is one small part of the biochemistry, but absolutely, yes. In truth, you can craft the exact flavor of cheese you want by balancing make parameters with strains and their endopeptidases (for most lactobacilli). Even possible to screen for strain characteristics through DNA markers.
What I am now interested to know is how people/companies decide which adjuncts to use
This part is mostly art, driven by a solid scientific backbone. Isolates are screened for properties that are desirable, then banked, then sometimes enzymes are extracted for a model cheese, or sometimes cheeses undergo prototyping in small form factors. You also have guys like me, and way more geeky than me, who can take a theoretical model based on enzyme types and predict flavor formation for a cheese, or at least predict sensory attributes. It's quite a bit of R&D.
combinations and quantities
This part is easier. You figure out your cheese attributes in the model cheese, and do predictive/combinatorial analytics given the attribute constraints and the enzymatic reactions. Most often, it falls in a few categories. Backbone for acidification, and enzyme types for texture and flavor. Where it becomes very tough is predicting synergies for flavor balance... and trial and error takes care of that.
It even gets murkier when I realised that each lactobacillus can have different strains which can result in different flavour profiles.
Doesn't stop there, they can mutate, and you have many co-factors influencing acidification rate and byproduct (therefore flavor) formation. For example, PAB-LAB interactions, or S thermophilus/Helveticus interactions. Basically, waste products from one can feed another, and you get into systems synergies. Moreover, food source and stress factors also influence flavor.
I guess it is just a matter of trial and error over time to see what flavour profiles develop.
There's quite a bit of that. But there are classic approaches, too, like aduncting DL with a single strain, or adjuncting cheddar with helveticus, or using veggie rennet and adjuncting with helveticus and rhamnosus to help with bitterness control.
Or is this information already documented somewhere?
Strain-specific details are available from the major houses. For this cheese specifically, I think it is a NIZO culture, or CSK, or related to it. Look up CSK (which you likely can't get where you are), they have flavor formation abstracted in this flavor continuum/wheel concept, and they've tried to use their strains to be like a dial. Want more sweetness? Add .2% specific acidophilus and you're there, want less back tannin in the alpine style? back off on helveticus, adjunct with rhamnosus, age cooler, etc. But the info you're looking for, that edge for exact flavor formation, it's generally not available to small producers. In a business where one single strain can account for a 5 million account or for a new product formulation (like parrano or zola), and where research and insight takes a great deal of money, there is little incentive to share key industry insights like that. The worst for me is when some arrogant jerk tries to blow smoke up my arse and doesn't want to give the most basic answer, like it's all some hidden knowledge.
I think you're on a decent start with what you're doing already. Taste it and see, and then tweak to get the flavor you want. If you taste and can't figure out what to do, post here, we'll help.