I believe that early positive column AC LPS lamps (like the Philora DA) had filament electrodes that were preheated before use, and possibly continuously heated during use. This obviously means that there had to be four feed throughs, two for each filament. Just like a fluorescent tube. No problem here.

But after after the Philora DO came out, they started instant starting the lamps. And this remained the case throughout SO, SOI, SOX, and SOX-E. But they never switched to one feed through per electrode, they always kept both lead wires. Looking at the construction of any modern LPS lamp, you can see that the two lead wires on each end are just connected together on the outside, rendering them practically useless.

Feed throughs are the one weak point, right? Wouldn't you want to minimize the amount of them that you have? Why not just have one wire feed through and attach the beehive cathode right to it instead of splitting it across two possibly problematic feed throughs?

With the technology sorted up well, feed-through reliability is not a major problem

Some reasons at least:

Manufacturing process certainly requires filament heating at some times for activation.

Two points is needed to support a beehive filament.

Two leads at each end are more mechanically robust. 

@RRK
Right... activation. Forgot about that part. Makes sense.