Natural infection produces a "better and more comprehensive immunity" for several reasons. But the main one is this:

Exposing the immune system to the whole virus triggers recognition of the entire portfolio of antigens and epitopes, instead of just the Spike protein. This means that immunological memory is more comprehensive and variants that my dodge Spike protein specific response (not just Ab but also CTL) would not be able to dodge the immune response to all the other antigens or epitopes.

I would like to add that immunity does not prevent infection, it prevents disease. What needs to be recognized is that re-infection is like a vaccine booster. It is actually good in one sense because that re-exposure is boosting and sharpening immune memory, including the maturation of higher affinity Ab.

Geert is saying that a natural progression of the virus in the human population will achieve an endemic equilibrium with a relatively mild (albeit nasty) version of SARS-CoV-2. In that natural equilibrium our immune systems will (by herd) make emergence of more lethal relatives unlikely. Those would be held in check if our immunological memory (repertoire of antigen-specific CTL, Th and B cell responses) were able to recognize the entire portfolio of CoV antigens. In contrast, herd immunity based upon the narrow memory (limited to spike protein) is vulnerable to immune escape with possibly far more lethal variants.

Dr. Noorchashm's argument is that a vaccine will reduce the total number of virus and that total number is the dominant factor determining the odds of a more lethal variant emerging.

This deserves careful thought. Which scenario is more likely to produce a high mortality variant - allowing natural infection to build natural immunity with a trade-off of high viral presence OR rapid vaccine-driven herd immunity with a trade-off that the immunity does not eradicate and is narrowly based upon certain spike protein epitopes?