The major takeaway for me is that failure rate is proportional to the square of ship size, but inversely proportional to the (non-squared) engineering tonnage.
Meaning that if you double the ship size, you have to quadruple the engineering tonnage to keep the same AFR.
Sure, but the same is true of doubling the number of ships. Two ships with 100% AFRs have a total of a 200% AFR. Increased failure numbers with increased tonnage is the expected behaviour, not a particular function of a big-ship design philosophy.
For me, what this really drove home is that AFR targets should never be fixed numbers - they need to be tonnage-based.