[nimih]

A 1% error rate honestly seems unacceptably high if you're in the habit of reading a large volume of papers, each of which has a large volume of citations. To quote TFA, "Science involves many different error-reduction practices that take time. It's part of the job." At the end of the day, it comes down to whether you think there's value in being correct, or if you're content merely seeming correct.

[dash2]

1% is my guess, and I'd call it a maximum. There is also a trade-off: when you use numbers, most people won't look up the reference at all. This means they simply don't know who deserves the credit. Or they may think a reference is more solid or accurate than it is.

[Ar-Curunir]

When I see numeric citations my error rate is much higher because I rarely go and check the bibliography because it’s such a pain in the ass.

Semantically meaningful citekeys give me useful information where I need it (in the text), as opposed to somewhere 20 pages away.