It might be a good idea to put a well matched 1 ohm resistor on each battery lead. But I really don't know what I'm talking about, so anyone else who knows better; please chime in.
The idea is to eliminate variances in the ESR of the batteries.
Okay, I will chime in.... First, I must point out I am merely a hobbyist. My solution has limitation and issues commensurate with my limited experience.
I have used recovered 18650 in parallel feeding a booster to feed 5V - typically use to recharge a phone via USB. My phone draw just around 1.1A max.
I considered the 1ohm resister solution and rejected it. At 1Amp, the resister will drop 1Volt. That is too much loss for me. Besides, it does not solve the issue of certain battery issues such as dead-short battery. Given my limited experience, I do not have a big mental tool bag to draw from. I decided to use SCHOTTKY diode. I choose 1N5817 because I have them and I like them (comfortable using it and familiarity with it).
My setup has two slots for 18650, but I can use just a single 18650 or use two. Each 18650 connects to
3x 1N5817 (3 diodes in parallel) to the "power bus" supplying Vin of the booster. (See attached photo, the switch controls if I am using 1x18650 or 2x18650 or off.)
Here is my thought process:
Since the diode has Vf (voltage forward is voltage-drop) that increase based on current, and lower Vf = less loss. So, if I use multiple diode to split the current, I will be dropping down the Vf reducing loss.
For simplicity, I start my consideration with 2x 1N5817 in parallel. Manufacturing variation means likely one will be better than the other. More current will flow via the better one (lower Vf). As more current flows, the Vf increases. At some point, it matches the not-as-good 1N5817. So, the 2x 1N5817 "self-adjust" letting more current go the lower Vf path, then the one with higher Vf kicks in to share the load. After some experimentation, I found 3x 1N5817 was about the best balance keeping Vf at around 0.27V to 0.3V (average). I do get more (smaller Vf) but not much more benefit going to 4x and beyond.
So, each of my 18650 flows to the power bus via 3x1N5817. Similar logic applies to batteries when the batteries are out of balance. The better 18650 gives out higher voltage and more current, so higher (3x diode average) Vf drop - until the voltage matches that of the lesser 18650 with its (3x diode average) Vf drop and it kicks in to help power the load. So more current is drawn from the better battery but eventually both batteries contributes. In most cases, the batteries became equal voltage in about 10 minutes.
With them connected via diode, current from the better one will not flow into the lesser one. I am not sure how often a battery can died and become dead-short, but even if it does, I need not worry.
While I used only two (18650+diodes) in parallel, the same method can be done for more (18650+diodes) set in parallel to increase the total capacity.
That is my solution I have been using for a while. Works well for me but I am not experience enough to determine if there is a deadly trap in there or not. May be this "design" will benefit your thinking process in some way.
Edit: Reworded a few places to added the words "in parallel"... The context is probably clear to most that the diodes are in parallel but may not be clear for everyone.