I wish I had read your thread sooner.
I like the new plan better because you can also more safely have the packs "prove" themselves with use by running the pack with only the ends of the strings paralleled. With that many packs you are almost certain to have some duds, or at least some with a significantly weaker cell. The reason I say it's safer is because If you force each pack at a parallel level to have an identical voltage, then a low cell has to be has its effect spread over the 4 other cells verses 19 other cells without the bus bar.
eg. Let's say there's a bad cell whose voltage ends up 0.5V low . That puts you overcharging the other 5 cells even if you were charging only to 4.15V . With only 2 bus bars, the effect is spread over 19 cells, so instead of a too high 4.25V charge, they end up a safe 4.18V. With that many packs you obviously won't individually balance charge each pack on each cycle, and if you go with full parallel structure at the cell level, the bad cell remains hidden. Cell level paralleling and monitoring is safest, but it forces you to test the packs under load before the pack is assembled in order to sort out the weak packs. I have no means of doing that, so I test with conservative real world use. After 10-20 shallow cycles, those which stay in perfect balance are trusted enough for cell level parallel structure and weak packs removed.
I've had a brand new unused pack flameout while sitting undisturbed for a month, so I treat them like fire bombs until they have proven themselves in use, so there's just no way I'm letting a weak pack hide inside my pack. Each has to prove worthy.
John