Some level of retroactive regulatory changes are required because the industry keeps discovering that the previous way of doing things was substantially less safe than previously assumed.

My favorite example of this is that during the golden era of cheap nuclear power mentioned in articles like this, it was the norm to run all the redundant control and monitoring wiring through the same narrow duct in a wall meant to stop fire spreading, fill it with highly flammable foam, and test the foam for air leaks using a bare candle flame. The way we learned this was a bad idea was because workers at Brown's Ferry Nuclear Power Plant actually managed to start a fire and take out a bunch of supposedly redundant monitoring and control systems whilst flooding the control room with smoke. This bad design made both the redundancy and the firestops that were meant to be there ineffective, and the stricter fire regulations required to prevent issues like this are a major cost.

You can't just assume that because something hasn't caused a major catastrophe yet that it's safe to continue doing either. This is such bad engineering practice and has played a role in so many major disasters across multiple industries there's even a specific name for it: the normalization of deviance. It's dangerous because it invalidates all the engineering and safety calculations that were meant to prevent disaster, replacing them with a gamble where no-one really knows the odds.