Nice!

I looked at the TPL5111 for #TritiLED at one point. I didn't go with it because the maximum frequency is too low to make the LED look constantly on.

It's a very good part for a flasher, though.  Good idea.

There is also LTC6991 that can go to 1ms, but with 2 orders more energy, same as sleepy micro category.

Ted 10 points for you, thanks.After adding best cap I had, epcos 100nF (B37987F5104K54), my meter seems to notice current spikes at ~4uA - consistent with TI Launchpad plot. As cross check, I made sweep with ref200 at 100uA and what I can read from plot, is ~0.5-1uA probe resolution and 1-2uA accuracy (assuming ref200 is spot on).So the spike on plot can be quantization errors in soft.Also if i remember correctly, probe has avg. for spike less than 1ms.

Based on that I had to assume no less than 1uA (max peak) current consumption for now. 

Interesting. Can you post some more details about the circuit?

The DMM appears to be reading 30na, but the scope appears to be showing 400nA pulses of close to 50% duty cycle, which would mean a higher average.  Unless I'm reading it wrong.

DMM's have difficulty measuring pulsing currents.

One trick you might use is to add a good-quality, low-leakage bypass capacitor (as large as you can manage) across the power supply to average the current draw.  This should bring the readings closer to the same value.