I don't know how closely you've studied the let's split design. To my basic understanding, each module has an arduino in it (they're like $3 each). When you plug the USB into an LS module it knows its layout and is just a small keyboard at that point. It listens for data through the TRRS. This data is just matrix locations, not keycodes. The "master" LS determines what keycodes to send using its mapping. Thus, you can switch layouts on an LS just by switching the USB from one half to the other.
For the USB splitting, you would replace the TRRS with a USB port and each LS module would emulate being a USB splitter with a keyboard attached (this is how Apple's aluminum boards do their USB ports and how USB boards can achieve more than 6 key rollover). Thus, each additional module just plugs into the end of the chain. From a firmware and PCB redesign standpoint this is actually your easiest path. Your big downside is that you lose modifiers across the boards - it's literally the same as just plugging in the modules all straight to the computer.
Pretty much any design you do, it's far easier to just have each board have its own arduino.