[mikeash]

You can run the first stage longer without any extra second stage fuel. Adding performance to either one (within limits) increases the performance of the system. Using the landing fuel for a launch would allow for a heavier payload without changing the second stage. The first stage would do more of the total work.

[mannykannot]

Ah yes, good point - I was implicitly assuming the liftoff weight was maxed out, by thinking that any upper stage weight increase would have to come from the 1st stage fuel load.

[tcheard]

Except that the longer you run the first stage, the longer you have to carry the weight of the first stage. It eventually becomes much more efficient to jettison the first stage and use a slightly less powerful but much lighter weight second stage, because you no longer need as much power to get through the atmosphere. Which is the whole point behind staging (https://en.wikipedia.org/wiki/Multistage_rocket#Optimal_stag...).

It turns out when you add 10 lbs of weight to the first stage for reuse you lose approximately 1 lbs of weight from the max payload. (http://aviationweek.com/blog/nasa-cnes-warn-spacex-challenge...)

On the second stage every 1 lb of reuse you lose 1 lb of payload, which is one of the big reasons why second stage reuse isn't really feasible on the F9.

[mikeash]

That makes no sense. It's never advantageous to ditch the first stage while it's still producing thrust. At some point when designing a rocket, it's better to add performance to the second stage rather than the first. But when presented with an existing rocket, you'll achieve maximum payload capacity by burning all fuel in the first stage before staging. Every second it burns is more delta-v imparted to the payload. The only reason you'd stage early is if you want to save fuel for e.g. a landing attempt, if the first stage is too powerful and the acceleration or speed would become too great, or if you simply don't need the extra performance to get the payload to where it's supposed to go.