This architecture would require you to monitor and balance each individual cell because the 99 cells in series in each brick all have a risk of being over-discharged (in top balance case) or over-charged (in bottom balance case). And given the strings are all independent (the only commonality is the overall pack voltage), that essentially means you have to monitor and balance all 6831 cells.
The assumed architecture where 69 cells are in parallel per brick means these 69 cells self balance and can be monitored as a single group. The BMS then only has to monitor 99 bricks, rather than each individual cell.
I agree with all that except that it doesn't require you to monitor each individual cell. Consider that it's possible to keep the cells in a brick connected in parallel most of the time to prevent the problems that you mentioned (over/under charge/discharge). It's also possible to remove a string from the ESS output if necessary so all bricks are affected equally if one of them has a few bad cells. But 90% of the time that will be unnecessary because you won't be discharging the battery very far. If the 69 strings are always permanently connected in series and used transistors/mosfets to connect them in parallel with the other cells in their respective bricks, it helps prolong and prevent the runaway failure.
I'm not saying this is correct. I don't believe any of us knows for sure unless you helped design the system. We always just assumed the cells in a brick were permanently wired in parallel. But there are a lot of benefits to having them permanently wired in 69 series strings.
That's why I want to know the pack behavior when Hansjörg's ESS began to lose capacity. It will help us figure out what's really going on inside it.