Or everyone else is getting it completely wrong, I guess.
That is still a large enough number to matter for a startup where those make or break growth & retention numbers matter. It would be insane to throw out 1% of your visitors while other employees are doing cartwheels in A/B testing for 2% improvements.
The general principle I follow is drop support if a platform no longer receives security updates. The second part of that is to continue to produce a simplified version of your site that should work on all browsers. This provides big benefits for disability access, low bandwidth users (which is all of the developing markets), and edge cases.
I disagree. I work at one of said startups, and we have dropped support for IE9 and below. We figure the 50+% increase in productivity (directly translated into new features, speed improvements, etc.) is well worth it.
I speculate that sites with low numbers of IE 8 users are typically oriented towards more technically savvy users, and/or if those sites were never working for IE 8 to begin with and so never had an opportunity to pick up IE 8 users in the first place.