Well to be fair to Enterprise Vault you have all the freedom in the world to set things up before archiving a single item, and as far as journaling goes, you can do custom filters and other third party add ons to journal items with different retention categories depending on who received the message or if it has a kind of attachment or keywords etc etc etc
It's why i'd advise anyone who is setting up EV to make sure they have their needs and future needs clearly mapped out and bring in consultancy as a misconfiguration at the beginning, especially with a compliance device such as a Centera, can cause major headaches in the future.
The idea though of seperating out the sites to me is a little extreme not only because of the sharing of items, but because you can potentially break compliance, for instance you cannot import a PST file through the VAC to a journal mailbox, you have to do it through EVPM.
However if you import the PST files, you lose everything that the journal task does, for instance Journaling will go through and expand distribution lists and what not to determine who receieved that email, but with a PST import it will only store the Distribution List
So for instance lets say i send an email to DL-IT-OPS, once its sent journaling will do the following
1. Open the email and expand out DL-IT-OPS
2. It finds that its been sent to UserA, UserB, UserC and UserD
3. It then adds that to the saveset and to the index
4. If i search for an email receieved by UserB, this email will show up
If I was to import the PST in to the archive and i search for email received by UserB, it will not show this email as being received, but if i search DL-IT-OPS then it will show that that distribution list got it, but not it's members
Also if i do have things like custom journaling and filters, this will be negated by a PST Import.
So the solution then sounds like you should import the PST files in to a journal mailbox and let the journal task take care of it. however you can then end up getting false positives
For instance lets say
2009 - DL-IT-OPS has UserA, UserB, UserC and UserD
2010 - DL-IT-OPS has UserA, USerB, UserD, and UserE
You then get one false negative and one false positive
Because EV will take the message sent to DL-IT-OPS, expand the distribution list and it says the recipients are A, B, D and E
So if you search for emails receieved by UserE, it will show this message, even though when it was sent, the user was never a part of the Distribution List
If you search for emails receieved by UserC, it will not show the message, even though the user did receieve it, they are no longer part of the distribution group so won't be registered as such.
Another thing also is that when you export items to a PST file from a journal, it no longer has the envelope message, which contains a list of all the recipients that got the email, including anyone who has been BCC'd
So what Enterprise Vault will do is see that the envelope shows it was sent to
UserA
UserB
UserC
UserD
It then opens up the main message and see's that it was sent to
UserA
UserB
UserC
This means that UserD is now an "Undisclosed recipient" and is essentially a BCC recipient, which is useful if you want to find out if anyone has been sending email "secretly" to someone who should not be getting this.
I really think scanners original idea of changing the retention, changing people not to expire, then expire the items and change the retention category going forward is the simplest and easiest solution that you have.
But again, if your centera is in compliance none of this makes a difference