I've solved it for myself, but it may not help others...
Essentially the problem is with the program at: /usr/lib/plesk-9.0/postfix-queue
This is the "hotfix" they keep replacing. Also the way that plesk is managing postfix mail processing is very inefficient. They fork/exec the postfix-queue program twice, and postfix-local once. The message passes through 4 or 5 processes before it gets delivered.
Since I don't use any of the spam, virus, or any of the extras plesk provides, I've totally eliminated them. I'm still in the process of testing though.
I've updated my /etc/postfix/master.cf to assure postfix-queue does not get called, ever. The two places it gets called is for 127.0.0.1:10025 (with before-queue processing) and 127.0.0.1:10027 (with before-remote processing). So looking for where mail is forwarded to 127.0.0.1:10025, I replace its "smtpd_proxy_filter" to go to 127.0.0.1:10026. So the following two lines are changed:
OLD:
smtp inet n - - - - smtpd -o smtpd_proxy_filter=127.0.0.1:10025
smtps inet n - - - - smtpd -o smtpd_proxy_filter=127.0.0.1:10025 -o smtpd_tls_wrappermode=yes
NEW:
smtp inet n - - - - smtpd -o smtpd_proxy_filter=127.0.0.1:10026
smtps inet n - - - - smtpd -o smtpd_proxy_filter=127.0.0.1:10026 -o smtpd_tls_wrappermode=yes
It seems to work, but I'm testing quotas, mailgroups, etc.
One thing I did notice which bothers me in testing quotas is that regardless of whether I bypass or allow the postfix-queue to be called, it actually queues messages that put a user over quota rather than bouncing them. It marks them with "Message can not be delivered at this time" and they remain in the queue (for 2 days now).