Is there any significant architectural difference between non-inverting and inverting tri-state buffers? In what cases would it be advantageous to use one over the other? To me they seem like the same thing just differing by an inverter.
Mostly, they are the same thing with or without an inverter.
Construction-wise, the simple buffer designs are naturally inverting, so the non-inverting ones have to have an extra layer of logic to invert the signal an additional time.
This leads to both larger silicon area and longer signal path through the inverter.
With both parameters historically being critical and with the inverting design often being slightly more versatile (you can use it as a buffer, tri-state, and/or an inverter), both options were offered.
One inverts, and the other doesn't.
You would use the inverting type if you needed the buffered signals to be inverted, otherwise you'd use the non-inverting type.