Thanks all, I have done a lot of reading based on all your tips.
I don't think using the traces for current measurement is a bad idea, but i woud need to do a lot of testing and redesign. I think it will be a little to tricky for me at this point, and i do have limited time.
So with everything i learned I think I think have the following option:
Hall effect sensorsI can stick with my ACS712/ACS758 sensors. However to get them stable i would need to redesign to a differential ACD setup, as i prefer not to use the Arduino's ACD's. A solution would be adding another ADS1015 and a voltage divider with either a potentiometer, or precision resistors or a software offset in the uC's code.
Current shuntThe INA219 is a great device. Extremely easy to use and seems reasonably stable. I would need 1 per channel as chips like the INA3221 require higher shunt voltages. A small advantage of this might be that I can place them very close to every shunt. A 0,0015mOhm shunt would be perfect. A single one will cover the 0-20A range nicely. A double one (On both sides of the PCB) will be perfect for the 0-50A range.
I made a table to compare the advantages and disadvantages. I hope i got the data right.

So looking at this, there seems to be some advantages to switching to shunts. I'm not sure how these translate into the real world
For now i wanna assume the 2oz traces ans the small SMD shunt are up to the task of transferring the 50A. The dissipation however will be a lot higher for the 50A shunts. I fear the PCB traces and components will already add more heat than i would prefer. Will the additional dissipation be a problem?
Also a 0-37mV signal is MUCH more sensitive for disruption than a 0,5-4,5V signal. How stable can i get the signal at a LSB size of 10uV? I gues i dont really need that level of precission, but 40uV/50mA would be nice to have. Do you guys this is possible, or will there be to much noise etc.