The "maxchildren" error message says that you are running out of
spamd processes -- there are more messages coming in than there are
spamd instances to handle them. I would suspect that the other problems are related to this. For example, the lint probably cannot gain access to the Bayes database if there are multiple
spamd processes in queue to access it, and presumably Sendmail is configured to return a transient error if it can't run the spam scan.
See if you can add more
spamd child processes, and/or make them go faster. If you are not using Bayes journaling and/or a database back-end for the Bayes data, those are some of the standard speed-up techniques for
spamd. You should probably visit the SpamAssassin wiki for more information, anyway.
PS. Your friendly Toyota dealer would probably appreciate if you didn't post their email address in the clear. That's how you attract spam in the first place. You can click the Edit button in the lower right corner of your post to go back and replace the email address with
whatever@example.com which is guaranteed by the IANA to never correspond to a real, live email inbox.