I think it’s really quite straightforward. Regardless of series or parallel the capacity would be used up equally, to within very small limits.
For information I believe they are in series, otherwise you would not be able to measure them as 3V+ at the top of the two batteries. The fact that they mount in opposite directions also gives an idea that points to this too, though it’s no guarantee.
If you ask me, your statement #1, that you dismissed, as highly unlikely, is the most likely, or at least a similar reason. These batteries are not military spec, they are $1 a pop, and they will not be made equally. Bear with me here, because this only makes sense as we get further into the explanation. Yes, they will be very close. They will use a method of manufacture that can be repeated to a high degree of precision, but not perfectly.
It is perfectly possible that one battery had 1% more overall capacity to begin with, and in normal use, mid life, this isn’t obvious, because the discharge curve of these batteries is very flat. I am 6 months in with my front door camera, and I just took it down to measure them, with the following results, whilst in the camera, and under normal rest load from Blink camera, in that motion detect was off, for the test.
Left battery: 1.692V
Right battery: 1.692V
Now I’m not using a precision calibrated voltmeter but that doesn’t matter, because we are doing a relative comparison. For argument’s sake, let’s say they did have a 1% difference, in overall capacity at the start, there is no way that this 1% difference would show up as a difference between the batteries, just because it’s 1% further down the middle of the very flat discharge profile. And, don’t forget, that 1% could represent many days of use, on these ultra low usage cameras. Roughly, 800 days, 1% would equal 8 days.
This situation changes dramatically though when we get to the point that the batteries are almost depleted. Once a given battery is very nearly empty, especially at low demand, such as imposed by Blink cameras, the voltage practically drops off a cliff, for only a few mAh of extra capacity being used. According to their own graphs, posted by Joel_Ek, further up the thread, this occurs incredibly quickly and if you look at the end of the discharge profile it is almost vertical. I include a link here for clarity.
Thus, one battery that gets there first, because they can never match perfectly, will be tanking down the vertical part of the profile, losing handfuls of voltage, in very little time, delivering very few mAh of capacity, whilst the other one, is still hanging on at the top of the cliff.
At this point, with one battery having tanked, the camera cannot operate, we no longer have the required 3.3V total, or whatever it is, so we replace them. When we replace, we check them, and we find one is half way down the cliff, and the other one is still at the top, or at least we would, if we measured them whilst they were still in the failed camera, under load. The issue is clouded somewhat by taking them out and letting them rest for a while, because this allows them to rebound somewhat, and you could find them both, again, at the top of the cliff, with similar voltages, but here the length of the rest period would be important.
Despite one possibly having a significantly higher voltage, I can almost guarantee that they will both be very close to being completely spent.
As for all being created equal, the remaining two brand new ones in an opened pack, just measured as follows, repeating it several times on each, to confirm it was a stable and repeatable result. Both had been left in the packaging, and had been untouched until now, and I handled them only by the sides.
1.797V and 1.798V
Again, the accuracy of the value isn’t important, but the relative value is, and they are slightly different, straight out of the box. Who knows what this means in terms of capacity, because the top of the graph is different again, in that there is quite a drop, initially, for little change in remaining capacity, but it does show that they are indeed not created equal. Very close, but not perfect, and that difference could easily explain one battery falling off the cliff first.
Who knows what their actual manufacturing tolerance is. Does 1.78 still pass? How about 1.7V? We don’t know, but we do now know there can be a difference. I do have a whole bunch, so I could test at least 10, to see highest and lowest, but I’m not going to open up unopened packs, just to see.
Anyway, it made sense to me, once I saw the graph, going from almost horizontal to almost vertical, in the last tiny fraction of the remaining capacity.