Comment # 7 on bug 1002492 from Jean Delvare

Having read the code, I now understand that the checkproc failures are actually
expected. Even if I delete agent.info, I guess the agent.info-endymion:0 check
would fail too after a reboot, as a new gpg-agent process has to be started at
that point. So most likely the real fix here would be redirect checkproc's
stderr to /dev/null so that the error messages don't get logged.

Why journalctl only shows the first error when I'm not root would be a separate
issue.