This is a bit of a stretch from what I said - 1% drop after the entire population has gambled through 10x their net worth is not meaningful. I also pointed out other speculative activities which we encourage, presumably because they compensate by growing the economy. Insurance might preserve or increase equality, but it also might extract so much rent that the overall utility is lower. There is simply no cut and dry explanation - for some parameter choices things work the way you say, and for some they don't

I think you are simply wrong about your assumptions. If the system is semi-stable the total utility won't decrease that much. If it's not stable it will. I am not sure exactly what you are missing there but maybe that it's expected utility going down, not utility going down with every outcome. For example if people with 80 and 120 net worth flip a coin for 20 it might be go up or it might go down but u(60) + u(140) < 2x u(100). Maybe that's why you utility model predicts the total utility collapsing quickly while in fact it predicts slow utility decrease (if the whole setup is close to being stable).

My model was not to demonstrate that utility doesn't go down ever, it was to show that it can do that extremely slowly, which makes the utility argument about why we discourage it societally a bit weak - we're clearly not very good at discouraging any other behaviors resulting in long-term bad outcomes (for society or the planet), and we reward all sorts of risk-taking.

I think the simpler explanation is that gambling is seen as addictive and destructive on an individual level, and there is no need for total utility to explain why that's undesirable