So I’m assuming these are WS2811/12, I suspect the principal is similar with SPI ones, but the encoding is different. In theory you can branch them as far as you want (assuming the voltage drop doesn’t cause artifacts, and if it does, you can patch in more down the line using fat wires assuming your PSU can handle it), the protocol doesn’t care.
The reason for this is because the way those work, each “frame” is a reset signal (long period of the data line being low) followed by a bunch of bytes encoded using a variable pulse time (similar to an RC servo, but differently shaped, faster, and digital, not discrete). Each light then sets its color based on the first 3 bytes in the stream, and retransmits the rest (with a reset in place of the bytes it just used), so the next light sees the second set of 3 bytes, changes its color, and retransmits the rest, so as the signal goes along the chain, each light chews off the bit of data it needs and sends the rest along.
The reason you can split them like this is because each light exchanges its data in the incoming stream for a longer reset period, so the signal is valid at any point in the chain, so you can easily just send one signal to 2 chains, and they’ll display the same thing, and because the output signal encoding is the same as the input (just missing 3 bytes), you can hook 2 strands up to the output of one.
If I ever meet whoever came up with this scheme, I’ll buy them a beer, or five, cause it’s awesome.