The concept has been tried a lot actually, just not in a complete fashion. The Saturn I and IB clustered Redstone and Juno (I think) fuselages in the first stage, although the engines were mounted seperately. Of course, the Delta IV is the most obvious current example.<br />Any time you cluster, you pay a significant weight penalty over a larger single structure. The reason to do it is if you either have a really good rocket and can't afford to design and build a larger, or for technical reasons can't build larger. The second reason affected the Soviet program. When they started working on the larger engines that were essential to their first ICBM (which is still the first and core stages of the Soyuz), they couldn't solve some of the problems with burning Kerosene and LOX in larger nozzles. So they clustered four nozzles together in each engine and never looked back! That's why the back end of the old "Semyorka" looks like a toy rocket; there's five of those engines, plus verniers (another older feature you don't see any more, like the old Atlas). The Proton rocket uses a type of clustering more like the Saturn I/IB, with the engines mounted seperately (those are not strap-on boosters surrounding a core). This was done because anything shipped to the launch site had to be brought in by rail. <br />Anyway, the idea of total clustering; i.e. mass-producing one rocket, and maybe just changing nozzles for altitude, as you said, is definitely do-able. I think Falcon is planning something like that. While it may not be as efficient, it could pay off, if the production process could be made less costly (trade efficiency in the factory for efficency in flight).