Let's say you start with a 4% SWR and a 70/30 portfolio. Over 20 years it goes pretty decent, your portfolio doubles in real terms, and you're down to a 2% SWR, spending the same as before (inflation adjusted). You do have enough to ride out bad times, so you can either go more aggressive, or less, or the same. I used to be in the camp of might as well let it ride, because you can afford to.
This is a nice summation of how the same set of facts can induce two completely-opposite responses, both entirely rational. That's why I find it weird that someone as experienced as Bernstein would come down so strongly on one side. It seems like a very "first-order" level of thinking to ignore the "if you've already won, keep playing since you don't need that money anyway" aspect. I think the very unsettleability of this issue hints at why asset-allocation/risk-tolerance will always be a very personal issue, and that despite all the math and research we can do to optimize various aspects of investing across the population, this is one area where there will always be multiple, divergent answers.
However, math and research can still help. Even someone who comes down on Bernstein's side and decides to "stop playing" can fuck themselves over.
Using your example, according to cFIRESim, a 2% withdrawal from a 100% bond portfolio results in 20% failure over 50 years, while a
4% withdrawal from a 100% stock portfolio results in only a 13% failure. Anything less than 20% stocks is more risky than a higher stock-allocation over that timeframe.
Or in another comparison, at an equal 2% withdrawal rate, a 100% stock portfolio and a 70/30% portfolio both have 0 failures over 50 years. But the 100% stock portfolio has a "lowest ending portfolio" value larger than the mixed portfolio, indicating it's actually the less-risky of the two un-risky choices.
I think the corrosive effects of inflation over long timeframes make the actual details of how to "stop playing" significantly less obvious for early-retirees than they may be for the too-rich old people that Bernstein is advising.