I'd like to not even touch this one but...I spent quite some time with Mitel trying to figure issues similar to this and it comes down to the MCD not monitoring networking trunks. I have around 45 PBX's...for example, in my corp offices there are 4 controllers, two controllers to handle users and two for trunking and each pair is resilient.
We all kind of gave up after about 3 or 4 months of trying to get this functioning correctly. Part of the problem that we came across is how all IP network trunk calls were marked as 9999, if my memory serves me correctly, so there was no real way to know where the call was actually destined to. Aside from SMDR collector instability, the SMDR collector service would stop after x minutes of not hearing from a site that may have lost connectivity. PF support claimed that it was designed this way..so often time we would find blank SMDR data files on the PF server.