When the battery gets low on charge the Roadster gives a "Reduced power" notification.
But recently I got reduced power without this warning. When this happened today, the battery charge was around 50 % (but the battery was -4 C cold, and -8 C ambient temperature). When I noticed this before, the charge has been lower than 50 %.
I think I found the reason: Brick #28 reached as low as 2.91 volts, it seems like a brick cannot go below this voltage limit when accelerating.
This seems to be how Tesla is voltage limiting the brick to preserve longevity. In the cell during discharge the positive electrode (LiCoO2) is decreasing and the negative electrode (graphite) is increasing. If the positive electrode drops a bit in voltage due to age and temperature induced overpotential, its not a big deal.In contrast, this could induce some deleterious reactions if the negative electrode raises too high. 2.91V for a short period of time should not be an issue at all. So overall, it is a good cutoff and Tesla is doing their job.
So getting to this particular case, hcsharp's question about the brick V uniformity is important to gauge health of the overall battery vs an isolated brick with higher impedance.
So why did this occur? Well, you could have healthy batteries and the combination of the cold temp of the battery, some growth of impedance of the cells with time combined with your low SOC may just have induced a dip in V in all the cells. Likely it was one brick (that's why hcsharp's question is important) .
Once the batteries are used for a bit, they will heat up and should have been normal.
This goes back to another interesting thread that was discussing an issue with peak power at normal ambient temps as cell age. Cells age two broad, but non independent, categories, true capacity loss (decrease in Ah, CAC etc) and also increase of impedance. The interesting case is the latter. For a while one could see no appreciable capacity loss but one could be power limited and see issues at low temp. For the former, one could have a decrease in Ah, but no appreciable change in power or low temp performance. They also show very distinct differences in eventual capacity fade. The impedance model is the one which , when capacity loss occurs, it starts to fail quickly as impedance rises above what is needed for your normal power requirements. We all obsess about Ah, but looking at V drops during load for each brick to calculate at least the high rate resistance with the known current would be very, very informative. Plot that for each brick as a function of age.. wonderful. True impedance would require electrochemical impedance spectroscopy, but if someone could write some code to extract the simple impedance mentioned above, that would be great!
Juts some opinions, I could be wrong.., at least don't stress about seeing 2.91V. Similar cells are cycled to many hundreds of full cycles at cutoffs even down to 2.75V, 2.9V should be perfectly fine for a short time. The cutoff is there for a reason. You just don't want to stop your cell at 2.91V and let it sit and risk bricking or pre bricking loss as the voltage degrades further.