Impressive commentary! Thank you for clarifying the triac methodology.
That was quite the observation for a 6 year old. I trust that was quite a
while ago...!!g So that speaks well of your memory, too. Seems you were
a "natural."
I have an attractive light fixture, which has been dark for years. The
fancy bulbs required for it burned out within days of
installation...ridiculous. Got tired of feeding cash into this fruitless
effort. There are four bulb sockets. The fixture is controlled by two
3-way switches; it's at the base of stairs.

This thread has renewed my interest in reactivating the fixture. I will put
a standard (non 3-way) slide-dimmer at each of the two switch locations.
SLIDE is critical, to force the user to warm the bulb filaments gradually.
The catch is that TWO DIMMERS cannot cohabitate in the same 3-way circuit;
because it seems highly imprudent to put two dimmers in series. Your
rebuttal would be welcome.
So...
I'll should split the fixture sockets. Two sockets will be controlled by
one slide switch; the other two sockets by the other switch. This way, at
least there will be some light, if only one switch is slid to its "on"
position.
I have another stairwell, which has two 3-way switches. One is a
push-button dimmer; of a 3-way nature. The other is a conventional 3-way
lever switch. The bulbs controlled there *never* burn out. I hesitate to
use this proven configuration with the fixture whose bulbs die so quickly.
I want those delicate bulbs to *always* light up slowly.

wrote
Your recollection may be inaccurate. Triac light dimmers actually work
by assuring AGAINST zero crossing startup. Except when full-on or off,
the dimmer operates by delaying the current with respect to the voltage
zero crossing, 120 times per second. When at 50% duty cycle (about 1/2
power), the current starts 1/4 cycle behind the voltage, i.e. at the
PEAK of the voltage sinewave, and stops when the voltage passes through
zero, so only the last half of each positive and negative half-cycle is
used.

Some heater controls modulate power by energizing the element for a
certain number of full cycles, deenergizing for some number of full
cycles, and so on. Here, zero crossing can be employed. This scheme is
not so good for lighting because the "blink rate" is well below 60 Hz
and would be quite noticable.

The question to be answered is how much of a difference does it make
using the last half of the sine "bump" vs. the first half, in terms of
energy imparted to the filament per cycle, taking into account the mass
and heat capacity of the filament, etc. etc.

I suspect that during rapid warmup, one segment of the filament gets
hotter than the rest, which makes its resistance go up higher, which
means it receives more power than than its neighboring segments which
still have lower resistance. Since it receives more power, it heats up
even faster and resistance increases even more. The hotter segment also
expands more rapidly and suffers greater mechanical stress and fatigue
as a result, becoming a likely point of failure.

Bringing the current up slowly, over the course of a second or so,
allows time for all segments to heat at closer to the same rate, as
well as time for heat to diffuse from one segment to another, allowing
more uniform heating and greatly reducing the chance of a local hotspot
forming.

Some of these ideas can be demonstrated by experimenting with bulbs in
series, which is effectively a single filament divided into segments.
Say you have a car battery (12V) and 5 to 10 identical automotive bulbs
in series. When you complete the circuit, you'll see that one gets
initially brighter than the others, then it may even get dimmer as the
others "catch up" to finally achieve uniform brightness. That's because
even though they are "identical," there are slight differences in the
bulbs and the one with the highest initial resistance and/or the
quickest heating filament will absorb most of the power till the others
heat up. I noticed this effect when I was about 6 years old, though I
had no idea what was causing it; I just knew I could rearrange the
order of the bulbs in series till they lit up from left to right and it
was pretty cool.

%mod%