Flaw with lava lamp entropy source

I've been running a 1970s-era lava lamp for some time, and found that it can enter a stable attractor where you get a non-circulating blob o' wax at the bottom. While Walker et al.'s (?) LL video entropy source is cute/clever, the general lesson we can take from this is to be careful that physical sources do not fail. Cooling the lamp and restarting it seems to have put it back into a quasi-random physical trajectory. I suppose my visual observation counts as an online entropic monitor that any physical source apparently should have. This was driven by a 40 watt bulb and the ambient temperature dropped when it stabilized. Shaking did not restart it; only cooling and then reheating did. Now back to your regularly scheduled war crimes.