So I realized (or re-realized, that happens often it seems) that dual swtich with OTSM has an issue. Eswitch needs 220kohms (ish) to keep parasitic drain down. OTSM needs max 4kohm (ish... but actually that high is already having a 20% impact on OTSM performance, would have been more without the powersavings option) to keep C1 discharge time fast for power-off detection.
These are directly contradictory. There are a few possible solutions:
1) Do nothing about it. you've got two switches anyway, just use the eswitch when walking around and lock out power with the clicky when done. A person is bound to forget this though. On the other hand, at 1mA of drain that's still 3000 hours off a 3Ah cell, which is a 125 days! So even if you forget, it won't drain overnight. I will drain significantly in a few weeks.
2) compromise, try to reduce C1 by 10 times, DEL doesn't like it, but it might work ok. Then you can increase the resistances by 10 times, 40kohm, so that gets you to 1,250 days... not too bad.
3) Need an active pulldown. This would basically require a depletion MOSFET, normally open, so when power goes down, it opens and drains C1 fast. It's an extra component on the board. I guess this requies another diode too, or some combined part. Hmm, seems there are ldo's with active pulldowns that could do it. This doesn't just replace the usual LDO though. There needs to be a barrier before C1 and one after C1.
By the way, this has nothing to do with the lexel option 3 solution. That's an OTSM firmware used without the OTSM cap.
Regarding OTSM, I ran my light down to LVP twice. No issues really. OTSM keeps working all the way to the end (2.7V LVP). I had long press set at 1.25s, so I don't know beyond that. I've changed the default to 1.5 since 1.25 feels a touch too short, and there's 3 or 4 s available.
The "worst" behavior I found was that when the LVP stepdown triggered it just marched its way straight down to off, when it seemed like there was enough juice to turn it back on in moon mode. I have trouble even caring. It was 2.8V and playing with it in different modes for about 2 minutes completely finished it off, so it's not like there was much left to save. Otherwise, no issues.