I think now that they have proven they can 'hit the hole', almost spot on, they'll be able to get approval to go for a try on cleared land space. It sounds like to me, landing on land would be ideal; less expensive and less risk with stabilization and rough waters.
Yeah, it sounds horribly unwieldy with a bunch of structural challenges, but the thing is it's self-stable in a way that a vertical column can never be. A few years ago the wisdom was that small (remote control) helicopters could only be flown stably with a coaxial main rotor. Quadcopters would work in theory, but everyone figured 4 motors on an frame was going to be even more complex to mange and that the stresses on the frame would likely weaken it too much, the power-to-weight ratios wouldn't work out and so on and so on. Now everyone is using quadcopter configurations because they are much easier to control and not actually that difficult to build.
My naive opinion is that output has never quite been consistent enough from rocket to rocket to make the idea of bolting multiple rockets together seem worthwhile unless they were all part of a single fuselage; one of the four would always be putting out too much or too little power and the whole contraption would fly off course in short order. but since the SpaceX rockets depend on liquid rather than solid propellant, and very precise pumping is not that big of an engineering challenge these days, it seems like it should be worth experimenting with on a small scale.