I cannot find the original post, but there is a known issue with some Stern ROMs where the Lamp Update code is too close to the zero-cross interrupt. If I recall, it only affects the first handful of lights that are commanded to turn on - as time passes from the zero-interrupt, rest of the lights are able to latch the SCRs and stay lit.

Meteor is a known example of that, and the Rocket 1 bonus lights.

The only thing you can do is add a lower-ohm resistor to the insets (I went with a 220ohm), which fixes the issue.

Edit: And darnit, I wish I could find the original post where this was mentioned.

Simplified explanation:
So SCRs work by 'latching' - once told to 'turn on', they will STAY on, until the current flowing through them stops.
This means that say, for example, if you told a SCR to turn on a light bulb, the bulb will stay lit - even if you stopped telling the SCR to turn on. And the bulb will STAY lit, until you pulled the bulb out (or cut a wire, the bulb burnt out, etc.) The bulb wll then stay off until you tell the SCR to turn on again.

Bally/Stern games operate by - every zero-crossing (that is, when the voltage of the Lamp Voltage drops to 0, because, AC wave), the CPU will tell the lamp board what lights should be on. This will happen about 120 times a second (100 times a second for those in 50Hz territory.)

Incandescent bulbs draw a lot of current (compared to LEDs) - they draw enough to "latch" the SCR on. LEDs to NOT. So, the flickering you see is the CPU telling the SCR to turn on - and then the SCR turning off immediately because there isn't enough current to latch it on. Adding a resistor - in parallel increases the current that flows through the SCR to cause it to 'latch'.

A 100k resistor will not be low enough - will not allow enough current to flow to cause it to latch on, sorry.

To sum up on the above - the reason it happens on SOME lights, and not others - like in Meteor and the Rocket 1 Bonus is because those lamps are the first ones told by the CPU to turn on.

If you look at an AC wave, it looks like this:
pasted_image (resized).png
When that line hits 0, the CPU starts telling the lamp board to turn on lights. The very FIRST LIGHTS it tells, the voltage may be a small fraction of full. (Say, off the top of my head, 1v instead of the 6.xv) Even with a parallel resistor, not enough current flows, so the SCR still doesn't latch. But as that wave grows up, voltage increases, the later lamps are told to stick on, and STAY on, because now there's enough current.

Note that the low voltage can cause this too - for example the same reason - if the lamps are only getting 5.3v, the resistors built into LED driver boards/adapters may still be too high to allow the proper current through to latch the SCRs.

Check this out: https://www.swtc.edu/Ag_Power/electrical/lecture/parallel_circuits.htm
It's how parallel circuits work.

The resistor should connect the two wires - on Meteor at least, one of the wire *may be* a braid attached to the foot of the socket. This is where one end of the resistor would go to. The other end would be the lug with the other wire.