This is a proof-of-concept of a 6-cycle loop.  It's conditional because 6Ting made it so I could only edit this in F43dit and it took 15sec for me to hit the play button in game.

I can for sure save:
  80g (16wsum) on trackloop arms if I 6T
  10c (10wsum) on latency from my sloppy conditional

I can probably save:
  25a (25wsum) by shoving the waste chain into the disposal too
 
So that would bring my best wsum to 1359-16-10-25=1308.

I'm considering some cheaper, smaller, and slower options, but we can see from this that they'd need a loop of at most 15.

Assume a cost >=500 (wsum 100) and area >=100 (wsum 100).  To beat this solution it'd need to finish in 1308-100-100=1108 cycles.  If we ignore latency, this would need a loop of at most 15 cycles.

Because of this, I'm going to explore 12 and 9 cycle loops, but anything else is probably not worth exploring.