Capability checks and run-time tests can only go so far. You end up with a bewildering array of code path fallbacks and polyfills that massively complicates the readability and maintainability of such apps.

I realize web developers think it's kind of marvelous and cool and philosophically pure to serve one page that scales from Netscape 1.0 all the way up to today's bleeding edge mobile browsers, but realistically, it is better to focus team resources on different tiers of functionality and in effect, write two or more difference pages optimizes for different scenarios.