"cooling-current"?! 

https://hackaday.io/project/162351-incandescent-ram/log/156134-relay-based-refresh-circuit/discussion-116477

You guys know how to keep me busy... got me thinking again...  How many stable states can a two-transistor circuit have ? How about four ? No clock signal this time... Just resistors and 4 LEDs that show the state.

Hmmmm, I'mma have to see this one. Be sure to link it here. Props, btw, for your 1-transistor 'memory' a while back!

The new 4-state circuit is found here:

https://hackaday.io/project/162405-2-transistor-4-state-flipflop

Slick

Indeed PTC devices are (or can be) two-state devices, like the polyfuses that are used for circuit protection (example PTS120660V005).

Problem with the bulb seems to be, that the filament is thermic isolated from outside world, by a vacuum around it. So when a very small current flows, the little amount of generated heat can not escape, and the filament will warm up, leading to a snowball effect. Perhaps cooling the filament would help ?

AHH. Indeed. Here I was thinking the isolation would cause it to be less-sensitive to such effects, finding its own equilibrium somewhere in P=VI. But, as you say, the heat can't escape. So, it'll just keep adding up. 

I guess, then, at some point equilibrium is found only when that heat can escape through other means; photons?

This suggests a highly non-linear resistance-power curve, highly dependent on time, etc! Especially at really low currents.

Weird.

My response, below, to the original comment grew too large for that thread. Great thoughts @DeepSOIC :

Interesting. It took me a few reads to grasp, but I think I get it. 

I vaguely recall having had similar ponderances long ago about... something, can't recall what.

The idea being that a device/component could satisfy Kirchhoff's laws two (or multiple?) ways depending on factors not typically considered. 

Here, a light being treated mathematically like a resistor, while really being more of a light/heat-producing thermistor, causes the circuit to have multiple stable solutions depending on the bulb's history. Yet, part of that, "history" is the actual "stable" state it's *presently* in, which is highly sensitive, then, to external factors; a cosmic ray might apply just enough extra heat to cause "thermal runaway" eventually causing the bulb/circuit to switch from one once-stable state to the next.

Two of these devices in parallel, driven by a constant-current would split the current 50/50 until a "cosmic ray" hits one, as it heats slightly its resistance increases, the other gets more of the current, causing it to heat as well. One might expect that to balance it back to 50/50, but the "damage has already been done", instead of balance we might get teetering/oscillation with ever-though-slightly increasing heat, eventually causing a switching of states, (plausibly from "working" to "blown")

For individual bulbs, though, there are multiple (long-term) stable states, no? E.G. if a 3V bulb is driven at 2V, it won't heat as much, so will be dimmer, and with a lower resistance. But, of course, at a different current.

That said, your idea still seems plausible, and if so, could allow for, e.g. "refresh" of an entire row of bulbs simultaneously by tying them in series and driving a constant current through. THAT'd be quite something!

And maybe more plausible being that these refresh pulses are just that, pulses, which might be too short to allow thermal runaway.

Hmmm... I needta get some paper!

Hmm, another thought: put a resistor in parallel with each bulb. Say a 'cold' bulb is 10ohms, and 100 when hot. A 50ohm resistor would draw current away from the hot bulb and push more current through a cold one. Erm, no... that's the opposite of our goal. but a parallel PTC... hmmmm

LOL wikip;edia: PTC thermistors 'latch' into a hot / high resistance state: once hot, they stay in that high resistance state, until cooled. The effect can be used as a primitive latch/memory circuit, the effect being enhanced by using two PTC thermistors in:;;; series, with one thermistor cool, and the other thermistor hot.[7]