1) legs are heavy
2) empty rockets are stronger in tension than compression
the scale of Superheavy is such that the above two items are making the arms scheme make sense. The number of engines also gives this rocket the ability to hover, which probably makes the scheme easier to pull off.
This was also true for this engine's predecessor, the Merlin 1D. However that engine's rocket (the Falcon 9) can't hover, as the power of a single engine throttled to its lowest setting still overcomes the weight of a nearly-empty booster.
All high performance engines tend to only throttle in a range in the upper half of the engine's performance, the difference making hovering possible in this case is that the Super Heavy has _so many_ engines that it can turn off. This is a sort of secondary throttling, or meta-throttling, and the rocket can use the combination of engine throttling and engine-off to hover comfortably (while near-empty) with three engines going.
Spent rocket stages are empty of fuel, but not necessarily of the ullage gases, the pressure inside could be e.g. 3 atmospheres and that could be enough to provide some stiffness in the direction perpendicular to the axis.
The engine load is probably a steady, consistent magnitude. While a landing load is rapid and variable. Also, you need to design legs for wind loading after landing, which can be high if you want to launch often.
the scale of Superheavy is such that the above two items are making the arms scheme make sense. The number of engines also gives this rocket the ability to hover, which probably makes the scheme easier to pull off.