Perhaps the emergency communication with satellites is a handshake, and the "test" mode for the beacons simply doesn't complete the handshake?

Possibly (the page I linked even mentions a "test" type of signal), but given that there is no return link for most PLBs and no "list of recent successfully received test transmissions" online, I don't understand how one would actually verify success.

This policy document mentions that the "self-test" feature actually does not communicate with satellites at all, and also mentions a "test frame" that is discarded by satellites, as well as a "test protocol" that is forwarded by satellites, but discarded by the ground segment: https://www.sarsat.noaa.gov/wp-content/uploads/POL-MCC-051-v...

So it seems like both the "self-test" and the "test frame" don't tell you a lot about how well your beacon would work in case of an actual emergency (unless you have equipment that can receive and interpret that signal), and live testing involves a lot of paperwork.