I think the more important aspect of his point is that there is very little data supporting those small margins. In addition (because of or regardless of data), a lot of people aren't comfortable doing those margins with CT-based imaging. You might be, but if the people recruiting and enrolling to that trial are not comfortable then the trial is DOA. Features of the environment (who is recruiting/enrolling, competing trials, funding, politics, etc.) have a huge impact on design in real life.
If you personally are comfortable, great. At this point, you can do a very easy QI study that is likely IRB exempt. There are many possible designs that would add knowledge, so you can even do it without an MR Linac. There seems to be a lot of people that are "sure" about the results of this hypothetical trial but no one is designing and running it as far as I can tell.
Don't get me wrong, these are definitely valid critiques and great adult discussion as opposed to Boomer/Lou tantrums. The critiques just seem blind to the realities of practical trial design, especially for a single-institution IIT that is internally funded.
As an aside from all of this, I keep thinking about how the Calypso folks must be kicking themselves haha.