In article , dpb wrote:
Don Klipstein wrote:
...
If the fixture is grounded, then the "electric field" ("voltage
gradient") within a bulb that has voltage across it but is not yet
conducting, is concentrated in one end around one of the
filaments. This increased concentration of "electric field" helps the
gas in the bulb break down there. Once the breakdown begins, the
electric field distribution changes, resulting in concentration of
electric field at the tip of the breakdown region - provided the
fixture is grounded.

Just how does it decide which end???

The bulb isn't tied into the ground, only supply/neutral.

In most 120V North American fixtures, one end of the bulb (or the "low
one" of a "series pair") is tied to ground via neutral if the wiring
diagram on the ballast is obeyed, even if the fixture is not grounded. In
USA, 120V circuits have "neutral" being a/the "grounded conductor". (And
the "safety ground" is the "grounding conductor", usually tied to the
"grounded conductor" at the breaker box, and "hot" is the "ungrounded
conductor" with full 120V with respect to "ground".)

The other end (of the bulb or of a series pair thereof) gets full
voltage of the line and any "inductive kicks", or full voltage of a
voltage-boosting ballast (such as the 120V-North_America-"traditional"
dual-F40 "rapid start magnetic ballast" and North American 120V "trigger
start" ballasts).

Grounding the fixture means that until the bulbs start conducting,
assuming their surfaces do not conduct along their lengths due to
hygroscopic dirt, that the electric field gets concentrated in the gas
around the "ungrounded end" electrode, and such concentration of electric
field gets the gas in that region of the bulb to "break down".

- Don Klipstein )