The top part I can't agree with, since there are already systems in place that we've seen that allow you as the player to control your ammunition expenditure to one degree or another. So it's a pretty clear misrepresentation to imply they've left it all up to the roll of the die. You have choices. You can use them. In my own designs for similar systems I have allowed troops of certain training levels to be more selective in their fire (either stingy with ammo or let-it-rip, with corresponding effects to both the to-hit/to-suppress odds and the ammo usage) and I hope that FL is tracking along those lines.And these simple and self-evident, yet crucially important, factual circumstances disappear completely in the typical Ammo Die system, where all consumption is abstracted into a single probability-based representation.
I did say "the typical Ammo Die system" ... it is true that what we've been told about the WIP system so far does seem to have more choice and complexity. But the thing is, it still yields unreasonable results, partly in enabling hero mag's — which of course is fine in a more Hollywood-esque setting, but to me, just feels wrong in T2k. (Still curious to hear a good rationalisation for how a 10-round magazine every once in a while can turn out to last for 11 ... 12 ... 15 shots, btw...)
And partly in that while you'd realistically know you have five grenades for the M203 — you'd really keep track of something like that, and you really do only load and shoot them one at a time — an Ammo Die system doesn't allow you to "know" anything of the kind! You have "d6" [or whatever] grenades left, and that may turn out to be any number of shots between two and six, or eight, or more, with a certain statistical bell curve.
It removes player agency in planning how to deploy that ammo, or give two to another character and still know you have three left...
So yes, a more complex Ammo Die system does allow more control over ammo expenditure. But it still, like all other attempts to reduce simple, quantifiable processes to statistical probability-based models, inevitably gives weird results at the fringes. Modelling complex processes that have or seem to have a large random element — like weapon hit probability — with die based systems is certainly appropriate, but the random element in how much ammo I have for my M203, and how I consume them, is relatively speaking minute.
There certainly are situations and circumstances that can introduce a greater random element to ammo consumption, as has been mentioned here and there on this thread, but to me, a good system would introduce random outcomes as and when those situations and circumstances arise; not have them be the status quo.
Under normal circumstances, a magazine isn't a "Schrödinger's Cat" box, where the content is impossible to know, until you actually try to use it, thereby collapsing the random waveform.