Just before the conclusion he addresses that; there was a revised/improved version in 1997 that was a commercial failure. Presumably due to the 555's ubiquity.

Analog chips can be real cash cows. Unlike digital microprocessors, a successful design can last for decades without any change in process node. You still get some really old process nodes being used on the analog side because they're still functional today.

Power applications in particular need larger transistors for sourcing and sinking large currents. Smaller process nodes don't help much for many of those applications.

Well, there are lower power CMOS versions of the 555 available.

A 555 timer chip can deliver 200mA at the output, so it can drive small lamps and relays directly. The output load usually dwarfs the current used for timing. So, for most applications, a lower power version is pointless.

I doubt that most applications of the 555 are driving relays and lamps, it probably is mainly used to generate timing signals. But that's besides the point, the low power versions can function with low voltages, allowing to power the chip with a single 1.5V battery, which would be impossible with the standard version, so it's far from pointless.

The CMOS version helps quite a bit when the 555 is configured for monostable or bistable mode.

Any re-spin will cost money (and a lot of it). As technology improves, the old processes usually become cheaper too.