Where was the Roman Army in AD408?

[Robert Vermaat]

Which is absolutely contrary to the primary source evidence, namely the Notitia. The West still had tens of thousands of men available as field forces.

IF all units mentioned in the ND were at full strength, indeed the army in the West would have had tens of thousands of troops available. IF.
However, you misunderstand the nature of the source. Whereas Zosimus or Ammianus mention the number in a battle, an army or a unit, the ND is a list of commands and never mentions numbers. It also does not mention if these commands were even manned by more than a skeletal staff, or even if they were manned at all. The ND is not a roster for the ministry of defence, it's a blueprint for the administration of the state.
Therefore it's impossible to take the list of units mentioned in the ND, combine that with a theory about general unit strength and combine both to arrive at a number for the Roman army based only on those two elements. That is the nature of the source evidence.

---

I suspect that the cumulative effect of this, withdrawing men and units, is why there are few troops on the Rhine in the Notitia.

And why most of these are not the units you'd expect, but mostly units named after their towns/garissons/regions, and therefore most likely limitanei or even local militias turned into army units.

---

I think Elton is maybe erring too much on the 'Roman' side in his analysis.
[..]
In fact, Elton's 'probable Roman' soldiers might more accurately be called 'indeterminate' - which would alter the balance of 'Roman' to 'Non-Roman' considerably...

If I recall correctly, Elton was primarily concerned with a reaction against the (Gibbon-inspired) opinions of historians that the Roman army was 'barbarised from top to bottom', painting a picture of bearded uncontrollable warriors (my exaggeration) against the former disciplined Roman legions. Elton realised full well that the evidence was scarce, but based on that he was of the opinion that no historian could make a hard case for a barbarised Roman army. Of course names are no hard evidence either, and under every Flavius there could be hiding a Gunthamund. But on the other hand, we don't know that either, and the fact that Gunthamund even bothered to call himself Flavius would for me be a good sign that the army was far more Roman than some think it was.

Which sounds rather like the situation in the late Republic, with Caesar's legions...
[..]
This 'regular Roman army' seems to be a bit of a chimera, appearing at one moment and vanishing the next!

I think your Caesarian analogy is a good one. The 4th century had seen an army and even non-Roman federates who had been loyal to the ruling dynasty, and this model had vanished by the time we enter the 5th century, changing to personal loyalties. When the troops mutinied against Stilicho I doubt that these men were only the Roman-born soldiers, but the majority of the units who have fought endless battles and may have looked for an easier future. I'm of the opinion that Stilicho was far too ambitious and fought wars to gain power over the Eastern Empire, which brought on the rivaly with Alaric and very possibly the mutiny of the army.

Yes, the defeat of Maximus, and then the very gruelling two-day battle at the Frigidus, must have severely depleted the field armies of both Italy and Gaul. They're both presented as victories - for Theodosius, of course - but could equally be seen as massive defeats for the military manpower of the Roman west, which may never again have recovered.

Theodosius took the best units that Maximus brought along and defeated what Arbogast brought later. Indeed that may have been the end of the army of the West as a campaigning force, no matter the units still on the borders. But the revolt of Constantine III may have ended much of that as well.