By careful selection of metric goals, impressive results can be achieved with little change. (this is a summary of modern american corporate process LOL). So i

By careful selection of metric goals, impressive results can be achieved with little change. (this is a summary of modern american corporate process LOL). So if your metric is pounds of payload delivered to orbit, for example, rockets are usually pretty heavy compared to the payload delivered so a rather small improvement in thrust to weight ratio can have seemingly extreme improvement in pounds delivered. You're still launching a million pounds sitting on the pad, its just removing a mere 300 pounds of insulation over than million pound booster means you get 1300 of payload delivered instead of 1000. Rockets have all kinds of interesting problems with thermal limits and combustion instability and its very easy to defeat those problems at considerable weight penalty and then when telemetry consistently shows the nozzle temp is fine at full flow rate and projections show 80% flow is fine, then next model has microscopically smaller pipes or whatever. And repeat many times and next thing you know you've gained quite a bit of performance. Its hard to add fluid flow baffles to an existing system, but if it works with existing baffles its usually much easier to slowly remove them. There is another spacex specific issue that to do the "land on a barge" thing general opinion is they overspec the marketing numbers to about 30% less. I don't know how much, if any, current improvement is lowering that to 27% or whatever. If you were to give up on recycling boosters and one-shot the booster, the marketed performance specs would go up a bit without any engineering changes (well other than software to land wouldn't be necessary, LOL).