To put it more simply still; AC light dimmers work by 'turning off' the sine wave part way through the cycle. This works on purely resistive loads (light bulbs) and blows up motors (ceiling fans).
LEDs are direct current devices that run on less than 2V. So they convert the electricity coming in to DC, then regulate it down to their voltage. No amount of 'sine wave chopping' is going to dim them.'
As a result they try to cheat, they try to both convert AC to DC and to measure the incoming sine wave and guess that it is being dimmed by an old style dimmer.
In the GP comment, the engineer chose perhaps poorly. Clearly as soon as power is available the light comes on full, then after the 'dim' setting is measured it adjusts the light level to the 'dim' setting.
What the light might do when it receives power is first do the measuring to look at the sine wave. then after it knows what brightness it is being commanded to show, then enable the LEDs. Now how long does it wait? If it waits 16mS (or 20mS for 50hz) then it can see a 'full' cycle. But since it could start anywhere it really needs to wait 24mS/30mS because that way it is sure to have at least one full cycle in its memory. However, if there is a 'smart' dimmer (think X-10) involved it may just pass through unclipped sine waves as it powers up because "the incandescent light takes many milliseconds to warm regardless." If the LED waits longer before it turns on you cross 100mS which is where pretty much everyone would note a 'lag' between turning on the light and actually having it come on.
All of that just to that a 20 year old dimmer based on TRIACs still works as 'expected' by the user.
Another lingo note, the old-school TRIAC dimmers are the forward phase ones. Being mostly on the commercial side I don't see much of this, but maybe on residential products that's what's implied when something calls itself "dimmable".
When we get 2-wire dimming it's usually reverse phase (AKA Electronic Low Voltage (AAKA ELV)).
To put it more simply still; AC light dimmers work by 'turning off' the sine wave part way through the cycle. This works on purely resistive loads (light bulbs) and blows up motors (ceiling fans).
LEDs are direct current devices that run on less than 2V. So they convert the electricity coming in to DC, then regulate it down to their voltage. No amount of 'sine wave chopping' is going to dim them.'
As a result they try to cheat, they try to both convert AC to DC and to measure the incoming sine wave and guess that it is being dimmed by an old style dimmer.
In the GP comment, the engineer chose perhaps poorly. Clearly as soon as power is available the light comes on full, then after the 'dim' setting is measured it adjusts the light level to the 'dim' setting.
What the light might do when it receives power is first do the measuring to look at the sine wave. then after it knows what brightness it is being commanded to show, then enable the LEDs. Now how long does it wait? If it waits 16mS (or 20mS for 50hz) then it can see a 'full' cycle. But since it could start anywhere it really needs to wait 24mS/30mS because that way it is sure to have at least one full cycle in its memory. However, if there is a 'smart' dimmer (think X-10) involved it may just pass through unclipped sine waves as it powers up because "the incandescent light takes many milliseconds to warm regardless." If the LED waits longer before it turns on you cross 100mS which is where pretty much everyone would note a 'lag' between turning on the light and actually having it come on.
All of that just to that a 20 year old dimmer based on TRIACs still works as 'expected' by the user.