I'm reminded of the old quip - the race is not always to the swift, nor the contest to the strong, but that's the way to bet.
As I mentioned here and elsewhere, most considerations of fault protection are a question of objectives, balancing the cost of doing something intended to protect against an expected failure against the lower cost of hoping that everything will work out OK and any failures that don't can be just absorbed. The easiest and cheapest up front way to deal with expected fault conditions is to ignore them. This works GREAT as long as you're willing to bury the remains and get a new on if one of the disastrous faults happens. This is actually an economic decision. If you're willing to let a given amp die or pay for the repairs, no failure protection is needed at all. If you (or the owner of a valuable amp) really doesn't want to take a chance on having to replace it, some money spent on protection might make sense. A sense of practicality has to rule though, as you're trying to say, I think. The most certain way to never have an amp fail ever again is to buy another one and never turn this one on again at all. Most people would agree that this was too extreme a position.
Rather than simply saying that the cost of protection other than >relying< on parts not intended for that purpose, a detailed analysis would probably come up with some way to evaluate what failures are too costly, and are worth spending money to prevent, and what other failures are trivial and don't need protection money spent up front, accepting the cost of repairs.
Let's look at that list and do some evaluation of what's being protected and what's being left open to repair costs. A good start with that is to figure out what's most costly to repair. In a given tube amp, the most costly part is usually the power transformer. The OT is generally somewhat cheaper, unless there is some sense of it being vintage/original/solid gold and too valuable to take a chance on. Tubes are wear items; they die in the normal course of operation over time. Money spent on protecting tubes in a circuit that doesn't eat up every new tube had better be pretty minimal. Resistors and caps? Well, maybe the main filter caps, but that's getting well down the list of replacement cost compared to the magnetics. The choke might be in there, maybe not. So let's look at the list.
a fuse on each HT winding phase of the PT
Probably worth a fuse. Those PTs are expensive!
a fuse between rectifier output and reservoir cap
Hmmm. What's this protecting? If it's the PT, better to just keep the two on the phases and skip the fuse. If it's the rectifier tube and/or filter caps, there is a real question whether to spend money on this fuse or just rely on the phase fuses. There are other ways to do this as well, not involving fuses.
a fuse in series with the HT supply to the OT CT
Protecting the OT from overcurrents? Probably worth the money.
a fuse in series with each screen grid resistor
Is this protecting the PT, OT, rectifiers, filter caps, or the tubes? If the tubes are "magic vintage" tubes, maaaaaybe? Probably not. Unless the tubes and screen resistors are too valuable to let fail, probably not. But you could if you just wanted to, I guess. If it's to prevent a runaway tube from killing the OT, a fuse in the OT CT is better.
and a fuse in series with each 6L6 cathode return to 0V.
The test again: what's that protection for? If you have the CT fuse for the OT's protection, you don't need a fuse per tube; tube shorts are already covered.
I count three fuses there. A fuse in the heater winding, two if it's got a hard CT, would probably do a good job of keeping the PT safe from faults outside the PT itself. So maybe five.
Yeah, it's a lot. But there are side-by-side phenolic blocks of five fuses available.
Taking on a couple of the precepts:
in compliance with the policy to only use something 'designed for purpose' when a protective function has been identified as being potentially beneficial, fitting a dedicated fuse,
Using something designed for purpose makes sense in any situation where you want control of what happens, not wishful thinking or hoping for lucky coincidences. And, as mentioned above, whenever "a protective function has been identified as potentially beneficial" isn't always the time to spend money, time and space on protection. Sometimes the sensible thing to do is to self-insure and count on simply doing the repairs. My curmudgeonly view is that if it's worth spending thought, time and money on protecting, (and even then, usually protecting against collateral/chain failures) it's worth doing it right.
Whereas there’s a long history of 1W screen grid resistors blowing in response to a screen grid short,
As above, what is the screen grid resistor opening protecting? If it's the OT, a fuse in the CT is a better place to spend your effort. The tube is already dead. If it's the PT, the phase fuses are a better place to spend your effort. Then there's the question of coverage. Will the screen grid resistor open in time to protect what it's supposed to? What if the last ones you got are just extra tough and don't give up in time? That long history is a collection of anecdotes. Sure, it happens. But stories don't get passed around about the ones that didn't protect by opening. If you're spending time protecting things, it's worth some thought about whether it will work correctly in which cases, what the effort is being spent for, and whether there's another way that provides more fault coverage.
a somewhat shorter but a ‘seems to work’ evidence of ~1/2W 1 ohm MF cathode resistors blowing in response to some power valve failure modes,
Same comments. It's good to ask (1) what am I protecting by doing this and (2) is there a better, more comprehensive and predictable way.
and the ‘but it works’ evidence of beefy high voltage diodes protecting the HT winding.
Sounds good - I've always thought that adding higher voltage, higher current solid state diodes in series with the HT winding made a lot of sense.
And surely everyone here has experienced the scenario of an actual, seemingly correctly specified fuse somehow hanging on and refusing to blow, despite it passing fault current that caused damage to the expensive thing that the fuse was supposed to be protecting.
That gets me back to that "but that's the way to bet" thing. Which is more likely to happen - a properly specified fuse not blowing or a resistor with no specification on how it behaves when it gets how much overheated opening at just the right time?
I’ve too have had ~1/2W 1ohm MF cathode resistors that have blown in response to a 6L6 shorting.
I’ve also put F330mA cathode fuses in another 6L6 amp, only for a short that developed in one of the 6L6 to damage the OT primary winding, before the cathode or T500mA HT fuse (yes I like fuses, honest) managed to blow and save it.
That can happen. But fault protection is a game of statistics, as you may have picked out of my comments. It's picking out "that's the way to bet", and deciding where to spend your time, thought, and money. There are anecdotes and counter-examples to nearly every principle. But one counterexample doesn't mean you should stop evaluating what's most likely to happen.
BTW a fuse in series with the HT feed to the OT CT is a bad idea; if it blows, eg in response on one valve shorting, the remaining power valves will then draw a very high screen grid current, probably resulting further damage.
Goes back to that "what are you protecting" thing. I agree that high screen currents can cause further damage. The question is damage to what, and what's the cost of protecting what's likely to die balanced against the cost of repairs if you don't protect, spread out by the likelihood that the probably further damage occurs.
This is worth digging down into the consequential failures a bit. The OT CT fuse opens. The screen grids then draw a lot of current. First up failure is likely to be the tubes with the high screen grid currents. A decision is required - do you want to spend money, parts and time protecting the tubes?
Maybe. If the OT CT fuse has opened, the likeliest cause was one of your tubes failing shorted. Possibly a wire touched the chassis, possibly the choke shorted internally. There is a string of other lower probability thing. But likeliest, one tube is already dead. That matched pair or matched quad is now unmatched, and that can't be repaired other than by replacing all of them. So it may well be reasonable to protect the remaining tubes, especially if you individually bias them to save on matching. The failure mode on the tubes is likeliest to be melting screen grids from the high screen current. This might be a place to do one of the following:
> take it on the chin and replace the tubes
> hope and pray that the small-wattage screen resistors you decided to rely on happen to open in time
> run all of the screen grid resistors to a fuse (Yikes! A fuse!) to interrupt the screen current
> use some other electronic means to shut down the B+; this is a favorite of mine, but outside the scope of this post
It isn't going to kill the OT, that fuse has already opened. It isn't going to kill the PT, there are fuses for that. It isn't going to kill the rectifiers; they're specified to take the current up til the PT fuse opens. It probably won't discharge the filter caps fast enough to damage them.
What else is likely to be damaged? I'm sure I'm missing something. Help me here.
And dedicated cathode fuses are kinda pointless, as they only protect against a certain valve failure mode, and are of no benefit for the shorts that send HT current via the heater circuit.
I agree. I'd do that protection another way. Again, the questions are what is being protected? In this case, it's probably the PT and rectifiers. A PT with phase fuses won't die from that. A rectifier tube might get hurt. But with auxiliary high voltage/current solid state rectifiers, the PT again won't take the hit. Filter caps will probably live. The heater being shorted to probably won't notice the extra current, and a cautious person might have heater winding fuses in any case. The same issues arise: what is already dead, what is being protected, and what is the cost and practicality of the protection weighed against the percentage likelihood of the failure happening versus the cost of repair.
And we know that the latter failure mode is a fairly common one, from the regularity of amps that have blown heater balancing resistors turning up on the repair bench, right?
Got any numbers on that? What's the actual statistical likelihood of any given tube failing shorted from HT to heater? It's noticeable when it happens, OK, but what are the real chances? Numbers matter.
Fault tolerant designs can easily be twice as expensive as a non-protected design, in which case a gig player probably ought to just buy a second amp. 
