I feel there's a paradigm difference here. Classical scientific method says "make hypotheses, test, confirm or infirm". That was the golden standard for 400 years. It works very well, but it has weaknesses (what doesn't). First is that it doesn't specify how you should select the hypotheses on which you want to perform (costly) experiments, and the second is that it's very bad at processing partial data. You can confirm on infirm - with some complicated math on how good the experiment was.

Nowadays we have much better protocols, mostly based on the Bayesian method which is much more robust on getting small pieces of information. One of its principles is actually this: no matter how small and useless an experimental result is, it should still budge your understanding of the phenomenon even if it's a tiny bit. There are some caveats here as well, but overall it's a whole different ballgame from the big, old, clunky scientific method.

So how does this apply to this situation? You have a cycle of 2 weeks for covid infections. A completely-pulled-out-of-my-ass protocol would involve 100 very-low-risk individuals split into vaccinated and control and infected deliberately with various doses. You get some data from this: how well is the vaccine working? how is the disease progressing depending on initial viral dose?

For the next 2 weeks you switch to slightly higher vulnerable populations, again split into control/vaccine.

For the next 2 weeks you ditch the control to lower the risk, and add another decade to the age of the population.

Now you have good enough data to start vaccinating vulnerable population at large.

You keep testing on volunteers, because you're about to vaccinate millions and you NEED extra data, like what's the fastest protocol to vaccinate everybody. Except now you have enough efficacy data to know that you won't kill your vaccinated volunteers, so you start using more mixed populations with higher ages - though the bulk will still be low-risk. And play with whatever variables you need adjusting, like vaccine dosage or booster timing, and adjust the population-at-large protocols based on that.

Compare this with what we did, which is literally last century methodology. We pulled out of the ass a protocol: vaccine dose of X, booster timing of 3 weeks. Then we tested it in a large scale long term trial. We confirmed it worked. And used it as-is, and declared that any change will require another large-scale long-term trial. Look at the definition of the scientific method in the beginning of my comment. It's just not common sense in a pandemic, except for bureaucrats minimizing their personal risk. Which is why my rather justified anger with the system.

Phase three clinical trials for the Pfizer vaccine ran ~16 weeks.

You need a minimum of 5 weeks before you can begin your challenge trials, starting from the same point, for the 3 weeks between doses and the 2 weeks after the second dose where immunity is being tested.

If you run three 2-week rounds of challenge trials after that, you're already up to 11 weeks to finish your challenge trials and begin administering the vaccines to the population at large, versus the 16 weeks for a regular-old phase 3 trial.

Now, five weeks is nothing to sneeze at -- given the other parameters, it's not unreasonable to think that on the order of a thousand deaths could have been prevented by using what vaccine supply we had available earlier -- but it's also patently not the case that we would be five weeks ahead of where we are now if we had started vaccinating earlier: we spent most of the last five months in the US supply constrained, and we didn't wait to start building out our capacity, so we'd probably be at the same point we are now in terms of total vaccinated today regardless of whether we started vaccinating 5 weeks earlier or 5 weeks later.

I get a strong "agile vs waterfall vibe" here, but I guess there's a good point. Russia did just that, i.e. skipped stage 3 clinical trials and progressively vaccinated volunteers. The EU was bashing them for "experimenting on their own citizens", but then had to acknowledge they should have been more aggressive with the trials.

At the same time, despite Russia's early start, they've now administered ~20 million doses, to a population of ~145 million, compared to ~50 million doses administered in the UK, to a population of ~67 million.

We have a natural experiment between countries that waited for standard clinical trials before beginning mass vaccinations, and countries that didn't. It's too early to take the "final score" [accurate, globally accepted final death tolls by country probably won't be available for a year if not years] to see who handled the pandemic best, from beginning to end, but right now it doesn't seem like vaccinating first made for a lasting lead in the vaccination race.