This article completely discounts existing install base as a viable market, as if hardware that was sold only a year or two ago is somehow nonexistent.

Check out how that same scary looking graph looks with a cumulative bar graph behind it: http://i.imgur.com/kjmnmRK.png

When you're looking at an install base that's nearly twice the size, that doesn't strike me as "too late" for anything. Sure, we can all read the tea leaves and want to get ahead of the curve and all that, but hitting the apex line on a tight curve is far different from just turning way too early and plowing into the guardrail on a straightaway.

Too late?

That depends on the size of operational overhead and opportunity of income.

If it's set it and forget it('ish), why not reap a few years income from a market that'll sell around 500M feature phones this year?

Taking into account that FirefoxOS is exactly targeting cheap smartphones which could completely kill feature phones, then yes, it's too late.