I think I found the issue... HEATERS!

[hbpowerwall]

Auto level still consumes a lot of energy across every cell bring them down not the low up (would use less energy)

 

[ajw22]

[...] I was thinking that lower capacity packs would bypass more since they would reach higher voltage sooner it would start to bypass. [...]

Nyes. That potentially happens only the first time you charge a not fully top balanced pack. Once the packs are fully top balanced, the lower capacity pack will drop in voltage quicker on discharge, and also increase voltage quicker on charge, thus converging at max voltage together with the other packs - no further balancing required.
On the other hand, endless continuous balancing is required if you have self discharging cells somewhere.

'fixing' them

Try my quickish yet foolproof way to find even the sneakiest of SD cells:
1. Disconnect the positive fuses of just half the cells, wait 30 minutes.
2. Set multimeter to the most sensitive setting (sub mV level), and measure the voltages between the positive busbar and positive of each of the disconnected cells. Good cells will read virtually zero difference, SD cells will stand out like a sore thumb.
3. Reconnect the fuses, repeat same process with the other half of the pack.

#1 cell.. ie Positive closest to the inverter so has the heaviest load

If you're saying that with your multiple strings of batteries, the closest pack has a higher draw... yes, but given your massive cable size the difference would likely be minimal. Would be interesting to see clamp meter readings.
If you're saying that pack#1 of any 14S system has a higher load, that simply is not true - physics dictates that all 14 packs will have the _exact_ same Amperes going through.

This weekend they are having a retirement party.

Are you going to just bin them, or test and reuse the cells? I know you have better things to do, but I'd love to see a post-mortem of the cells: capacity drop, rust, SD check, etc.