I can't help you on the specific BMS advice request.
If you go series-first and then parallel, in addition to more cell monitors (which do give you the advantage of watching every cell), you need a way to isolate each string. Contactors or FET gates are more expensive. Four BMSes (or a fancy BMS that can manage four strings) is more expensive. The wiring
to a central bus usually becomes a bit more expensive.
For this reason, many people do some degree of paralleling first. That's fine and justifiable. (Sometimes, they also use other technical justifications outside of cost that I am not sure anyone has ever actually substantiated. It's likely those folks will be around soon to espouse those ideas.)
The big benefit of series-first is redundancy and uptime. You can isolate part of your battery
or if an issue arises. Whether that is important to you is pretty situational.
"Safety" is a loaded term:
(a) From the point of view of a battery thermal event, I think either topology is about the same. LFP is very safe. Cell-level monitoring with HVC and LVC will work
fine in either scenario.
(b) If system uptime is critical, series-first is "safer." Someone with maneuvering thrusters tied into a battery system might care a lot about that. Or, in our case, we leave a dog on board unattended in places where reliable air-conditioning is important.
(c) In theory, a series-first design handles a single-cell hard short failure better, but in practice as a community I think we have basically zero documented cases of this ever happening. (Cells do fail, but in gradual and less exciting ways.)
Last thought on series v parallel approach:
Most people buying
drop-in (commercial) LFP batteries
today are in effect building a series-first battery. Each drop-in unit has an internal BMS and isolation mechanism. However many of them they buy -- and it is almost always more than one -- are ganged up in parallel. It is more expensive to do this. It works just fine. There are way more people doing this now as compared to us hobbyists building up big DIY
banks. I suspect LFP house banks are being assembled 100:1 or more as series-first designs now, because that is the only practical way for the commercial
providers to deliver scalable storage
But, horses for courses.