The major factor in Li ion degradation is largely calendar life time , ( they die just sitting there ) due to the constant parasitic action. After that high charge rates have a factor
I agree. But float charging them to the highest rated voltage for maximum capacity increases the degradation by some non-linear factor.
If negligible cycling happens, and floating causes damage mostly by calendar aging, then the function is indeed non-linear, and interesting. From what I have read and measured myself, it's clearly a piecewise function of two slopes:
Below about 75-80%, you have strongly increasing life expectancy, the lower you store it at. You don't need to go arbitrarily low; at around 50%, it's already so freaking good it's hard to say if there's anything more to gain.
But, contrary to intuition and common belief, over about 80%, strange things happen. The cells I have tested show very little difference at all (i.e., 80% and 100% are equally bad), and I have read a paper in which a Panasonic cell (IIRC) actually showed
better calendar life at 100% compared to at 80%. I really don't know the mechanism behind this (haven't looked at it). Do note that this is based on actual production chemistries available now (visible on both LCO and NCA), but is not a fundamental physical law, so if you read this post after years, things may be different.
Now, between 80-100%, charging current does more damage (than at, say, between 60-80%), so if a lot of cycling happens, then limiting yourself to 80% will be a benefit. But for storage, or very low-ripple, low-cycling floating, 80% is not any better than 100%. So you can as well use the near-full capacity; or go even lower to really increase the life.
You need to make the distinction between calendar and cycle life anyway: for calendar aging, lower temperature helps (the lower the better), but for charge current induced damage (cycling),
higher temperature helps (optimum can be near the rated maximum charge temperature, often 45 degC). (This is unsurprising; as you know, charging tends to be completely forbidden below 0degC. In reality, it's not a step function, and as there is no water anywhere to freeze, 0 degC is nothing magical, just a nice number where charging already produces "too much" damage.)
So when you combine both calendar and cycling damage (as you do in a practical float system), you'll have a combination of opposing constraints, and need to know which damage mode dominates. Even a guesstimate is much better than nothing.
For example, in a highly cycling system, I have measured that a
heated battery (at 50 degC) does over 1500 cycles just fine, but a cooled (at 10 degC) battery dies in around 100 cycles. The heated use case was outside the manufacturer's specification; the cooled one was inside the specification, yet died early.