From a dc network analysis approach, the only difference in a daisy chained and a star topology is where the connection points are made. It isn't as if a keypad has an rs485 in and an rs485 out. There just a connection point, with a cable coming in and out both connected at that common point.
Imagine you had 2 devices on a single branch, one 1000 ft from the panel and the other halfway between at 500 ft. Start moving the halfway keypad even closer to the panel... 100 ft, 10 ft, 1 ft, 1 inch. As that approaches zero, the connection point for that 1000 ft keypad also approaches zero, until finally it is actually connected at the panel as well.
As pointed out, it's a transmission line so complex impedance does matter, but for very short cable runs, like a foot or less in my case, that impedance is very, very small. Rs485 may be sensitive but not quite that sensitive. I guess my system is proof of that... although I was starting to wonder, it seems I was seeing just a programming glitch not a communications one, and it is functioning fine with 4 branches currently.
As a practical matter I'll fix that as soon as the hub arrives. It will make future expansion easier, but also if and when I run into any other issues I don't want to always be wondering if that tiny difference is actually making a difference!
