Why is Kr-85 used in HID lamps? I saw it on one of my Philips MV packages.
Maybe better starting?

It's used as a starting aid and it's a radioisotope of Krypton. See this thread for lots of details.

Thanks for the link! It's odd that it improves the starting.

Without it, adn in total darkness, the lamp is 2 metal electrodes in a chamber of inert gases, that are electrically not conductive. You may apply voltage to the electrodes, but why would any Electron leave the metal and go into the isolator ? The lamp would just stay like that, with voltage on the electrodes and no current flow

If the electrode is subjected to light, an incoming Photon got chance to kick an Electron from the metal electrode. And if at this time there is voltage present, this Electron which allready is freed from the metal will accelerate and hit the gas atoms and other electrode, releasing more Electrons and starting the plasma

But a lamp is supposed to be able to start in compete darkness too. And there is where the Kr85 comes in handy. The Kr85 emits particles as it decays, some of which hit the electrode and provide those freed Electrons, as Photons would

In some old FL starters that dont have the Kr85, they indeed have "fear of darkness" - If you switch on the light in complete darkness nothing happens. Then if you light a flashlight in the room, some Photons will get into the starter and start it up

Without it, adn in total darkness, the lamp is 2 metal electrodes in a chamber of inert gases, that are electrically not conductive. You may apply voltage to the electrodes, but why would any Electron leave the metal and go into the isolator ? The lamp would just stay like that, with voltage on the electrodes and no current flow

If the electrode is subjected to light, an incoming Photon got chance to kick an Electron from the metal electrode. And if at this time there is voltage present, this Electron which allready is freed from the metal will accelerate and hit the gas atoms and other electrode, releasing more Electrons and starting the plasma

But a lamp is supposed to be able to start in compete darkness too. And there is where the Kr85 comes in handy. The Kr85 emits particles as it decays, some of which hit the electrode and provide those freed Electrons, as Photons would

In some old FL starters that dont have the Kr85, they indeed have "fear of darkness" - If you switch on the light in complete darkness nothing happens. Then if you light a flashlight in the room, some Photons will get into the starter and start it up

Quite an easy way to play with physics.
But What If the Kr-85 decays enough to "go bad"?

The quantity there is tiny, but it is sufficient to put out a particle once in a while pretty much forever. And on average some part of the particles happen to hit the Negative electrode of the arc tube. As the material decays, the quantity of particles goes down (in half every X time, which is the half life interval of the Kr85). All this means is, that the lamp will wait longer from the moment of voltage applied untill arc appears, but it is still matter of a few seconds at most even for quite old lamp

Spent neon bulbs has this "fear of darkness" too. When they age they get flickery. In complete darkness it flickers at a slower rate. Shine a light on it and the flickering speeds up. Fun to "rev" the neon bulb with a flashlight.    

In many institution buildings from the 80s and 90s the lighting is controlled by a latching relay and push switches connected in parallel. In apartment buildings the stairs lighting is conrolled by a time switch, that is started by push switches (push for switch on - auto off after time is up), which are located in every floor of the building, in all utility/passage areas, and sometimes in the flats

Some of the switches are illuminated with a Neon and some are made of Plastic doped with Phosphorescent paint. Many of the Neon ones appear to light steady or with only little flicker when there is light around, but become flickery down to mostly off and only occasional flash when you cover them with your hand

Another fun we had as kids was to flash various patterns onto the glow in the dark switches (with a stencil and a flashlight) in the darkness, like a face or "contamination" pattern. Yours truly had also attempted Blacklight UV from a battery powered lantern, Photo flash, Solar beam, Hyper beam and the first Blue LEDs of the early 00's to make the switches glow exceptionally bright, to the point that they visibly illuminated the nearby area in complete darkness

What about using a dimmed incandescent to provide light?

That is used in the door bell switches at 8V..12V voltages. The door bell is designed so it does not react on the current the lamp works on, but rings when the switch is pushed and the lamp is shorted out. Could have used a LED, but apparently the Incandescent switches just do the work and nobody cares to make them with LEDs

On the lighting push switches at 230V its Neons, for few reasons :

 - they are very often installed in big clusters (10, 20 or more switches controlling the same relay), therefore the current draw of the lamp in each must be tiny so the combined current of 20 switches does not activate the relay when no switch is pressed. The Neon works at under 1mA, so many switches dont draw too much current

 - The switches light up all the time, so Incendescent lamps would cause quite big energy use. Especially in places like stairs, where the big lighting works only for short time but the switches work all the time, so the energy use of the switches might be not so small compared to the big lighting, if they were not the most efficient little lamps possible. For 230V and even 120V this can be only the Neons - at so low power levels, Neon + resistor is more efficient than LED + resistor, as the Neon provides same brightness at lower current/higher voltage, so the resistor for the Neon would dissipate less power, which is where most of the power dissipation of the indicators goes)

 - Making tiny Incandescents for 230V is not possible (what to do with all that long filament in tiny space), and using a resistor would be the most inefficient solution ever

That is used in the door bell switches at 8V..12V voltages. The door bell is designed so it does not react on the current the lamp works on, but rings when the switch is pushed and the lamp is shorted out. Could have used a LED, but apparently the Incandescent switches just do the work and nobody cares to make them with LEDs

On the lighting push switches at 230V its Neons, for few reasons :

 - they are very often installed in big clusters (10, 20 or more switches controlling the same relay), therefore the current draw of the lamp in each must be tiny so the combined current of 20 switches does not activate the relay when no switch is pressed. The Neon works at under 1mA, so many switches dont draw too much current

 - The switches light up all the time, so Incendescent lamps would cause quite big energy use. Especially in places like stairs, where the big lighting works only for short time but the switches work all the time, so the energy use of the switches might be not so small compared to the big lighting, if they were not the most efficient little lamps possible. For 230V and even 120V this can be only the Neons - at so low power levels, Neon + resistor is more efficient than LED + resistor, as the Neon provides same brightness at lower current/higher voltage, so the resistor for the Neon would dissipate less power, which is where most of the power dissipation of the indicators goes)

 - Making tiny Incandescents for 230V is not possible (what to do with all that long filament in tiny space), and using a resistor would be the most inefficient solution ever

Why not use a small LED?

In the Extra Low Voltage system (the door chime) no reason why not to use LED

In the Low voltage (120V or 230V) efficiency : The indicator must be as tiny and simple as possible, so its ballast cant be anything more than a simple resistor

With the Neon, about 60V go across the Neon and the rest across the resistor. So on average, the voltage drop on the resistor is 60V for 120V line or 170V for 230V line. This means resistor ballast efficiency is 0.5 on 120V or 0.26 on 230V. Not great, but for the little power of the indicator, fair enough

With the LED, about 2V go across the LED and the rest across the resistor. So on average, the full line voltage minus the 2V go to the resistor. This means resistor ballast efficiency is about 0.016 (less than 2% efficient !!) with 120V or 0.008 (less than 1% efficient !!) on 230V. The ballast to drive a LED in this application would be just epicly inefficient

Unless you wire lots of LEDs in series to boost the efficiency.
Maybe a Neon indicator, LED and resistor in series to also boost efficiency?

Available space inside the switch is very small. No place for anything fancy, and no need either. Its just an indicator that uses some 0.1W with the resistor

Available space inside the switch is very small. No place for anything fancy, and no need either. Its just an indicator that uses some 0.1W with the resistor

But it doesn't mean that LEDs can't be smaller.