Those of us who bother with accessibility have known since forever that some of the worst code comes from the Big Players.
How you earn your money aside, the assumption that code is good or best practices because “big players” and “large companies” are using it is a bad assumption. The only thing you can truthfully say about it is that their code and their models and their practices are working just well enough for them to manage to earn some amount of money.
That is not, and has never been, a benchmark for things like Good Code or Best Practices. I honestly don’t care if it’s good enough for your day job: the kinds of garbage I’m asked to copy and paste at my work is pretty bad, too. Doesn’t make it Best Practice.
“Amazon is huge, has a bazillion customers, and sells lots of stuff every minute of every day! Therefore, <i role=“checkbox”></i> and similar is Good Code and Best Practice”.
I honestly don’t know anyone (any developer) who would actually believe that statement. Instead, it would be
“Amazon is huge, has a bazillion customers, and sells lots of stuff every minute of every day! Therefore, <i role=“checkbox”></i> is managing to fundge its way into working for a certain (determined by the company) critical percentage of potential customers.” (at the time Derek Featherstone was testing Amazon, this fake checkbox was part of an extended order form. So if the user hadn’t finished a purchase, they’re still considered a potential customer here).
That’s all it means. It half works, or three-quarters works, or occasional blackouts and screwups or any other problems they have as a business are relegated to random little blogs here and there who get little publicity even when it’s outright theft (the link is not regarding code, but simply a particular business practice of a Large, Successful Company Who Manages To Earn Plenty Despite Doing Things Like This).
The “it works or we’d hear it from customers” argument has little meaning: customers are always complaining all the time about everything (legitimately or not), even holding to Will Hill’s 90-9-1 user participation principle, and while companies at times will make changes in response to significant (enough) complaints or money loss, this is not even the driving force for change within any large company, because of the costs involved in changing anything. Instead it’s usually marketing (including user studies, focus groups, introducing new or modified products), law suits, and bad publicity. Not in that order.