IMHO, on the linked page found here:
https://dev.mysql.com/doc/refman/8.0/en/osx-installation-pkg.html
Step 6 needs more information regarding the (apparent? alleged?) requirements of ancillary systems to use caching_sha2 (the screenshot provides more details than the write up).
In other words, if choosing the most secure option breaks everything, maybe "we" aren't ready for the most secure option to *also* be MySQL's default. The down-selection of a dreaded "legacy" encryption mechanism feels like I'm a non-compliant end-user trying to defeat best-practices, when in fact I'd much prefer to adopt SHA-256 hash.
Not sure if by design decision or deployment demand, but it seems as though this authentication component was well intended but poorly delivered.
Table 2.6 on the same page should also include users and permissions columns to greater assist troubleshooting. While wandering the 'net to find solutions, somewhere around 70% of all solutions pertained to folder/file permissions, often related to the _mysql user, which is not even referenced here.
Maybe the Internet has gone crazy, and Hillary Clinton has joined Donald Trump in obfuscating the _mysql user and its significance on a national scale. Maybe all those bloggers, git users and stack exchangers are all conspiracy theorists, and _mysql isn't even needed under 8.0.