Well, I fixed it today while waiting on a solution from our VAR and Mitel.
I won't post step-by-step since it's probably a one-in-a-million fluke to end up with a broken but fully installed cert, and if you're not comfortable enough on the linux command line and working with apache based web servers to get it from these details, you should probably get help form mitel and/or your VAR.
In any event, here are the parts that mattered:
-9.1 (and presumably 9.0 and future versions for a while) runs apache (or at least something that started life as apache).
-If you're used to working on linux web servers, be ready for everything to be in non-standard locations. If you're not used to working on linux/apache web servers, don't bother. Leave it to Mitel, and be prepared for them to say the best solution is to wipe the entire PBX and start with a fresh software load.
-'find' is your friend, although all the certs you uploaded will be not just renamed, but exported to different formats
-You can stop and start apache from the command line with"
service httpd-e-smith stop
...and of course...
service httpd-e-smith start
Interestingly, the following command errors out, despite 'restart' being defined in the init scripts.
service httpd-e-smith restart
If you've installed your own cert, the cert is stored in the e-smith home directory (because who cares about standards?) at /home/e-smith/ssl.crt/thirdparty.cer and the key is stored under /home/e-smith/ssl.key/thirdparty.key
It should go without saying, that touching any of these files is probably not covered under your SWA, and you'll be looking at billable hours to fix any of this.
As a final word of warning, NEVER enter the following command:
service --status-all
On every linux server I've ever touched besides this mitel PBX, that command is utterly harmless, and simply lists the status of all configured services. On my 3300, it got a few services in, spewed a mile of unintelligible error messages, and then unexpectedly rebooted the PBX. It took multiple reboots, a second attempt at running that command to see if it was a fluke (it was not a fluke) and ultimately letting the system sit after reboot for more than an hour before call controll came back up.