Another thing to consider is that for the mode spacing to look right it needs to increase by around 3x between levels. Just adding one 7135 at a time would look awful.
I had a flashlight that ramped (the original driver on a Yezl t9 before it broke). It was good for higher levels, but it was hard to select a very low level because they went by too fast. So how about a smooth logarithmic ramp, doubling the current maybe once a second? Of course there needs to be a separate way to access high.
Did anyone mention that forward voltage increases with (instantaneous) current, so reducing the amount of pulse width modulation needed would reduce the heat output of the LED, for a given average current, and cause the driver to stay in regulation longer, as well as increasing the light output?
The tint, on the other hand, may depend on temperature of the phosphor, so tint shift might also be reduced.
Radio frequency interference would also be reduced.
The only disadvantage seems to be the complexity.