I'll leave the specific technical answers to those who have greater practical experience in a corporate network envronment (have you already asked the UK support people for information?) but would observe the following.
The size of the reports, per se, may or may not be a significant factor. Up to about 250,000 records performance is pretty goos even on quite modest machines for a typical extraction. After that things may slow down a little but a lot depends on the what you are doing in the process. For eaxample building a lot of calculated fields will have an effect.
In terms of memory - the more the merrier and it is pretty cheap these days BUT you didn't say what your corporate policy is for operating systems and this will have an effect, potentially, on how much memory Windows can use. That said whilst one could see the potential for doing all processing in memory the system will still need to sync to disk regularly and so a fast disk would also be desirable. But all of this depends on what you are are consiering to be fast and slow.
So far as I am aware Monarch is not currently multi-processor/core oriented by design. However there are still likely to be some gains from using multi-core since I would imagine that you will have other processes running at the same time. Indeed just having the system processes running across the cores should be beneficial. Since multi-core and large memory capability (compared to the past) seem to be the standards these days it makes sense to go for as much power as you can get with one eye on probably future developments that will make use of it. The same goes for memory but at least that offers a more accessible and affordable future upgrade path should you choose to start with less memory than the machine is capable of supporting.
As a matter of interest, how large is 'large' in terms of the reports you are expecting to be be processing? Have you compared your expectation of report extraction table size with the technical information available in the 'Specifications' section of the Help file? It may give you some idea of how large your files will be in relation to design expectations of Monarch.
We use Monarch on a local network and all users have to have the directories mapped to the same drive letter or they do have to redo links or report files locations each time they open a project.
The only way you might get around it is if you have seperate projects for the off shore users that uses their drive letters. I have not tried it here but it might be worth a go.
That's an odd environment indeed, as the whole point of UNC is the "uniform" part.
But if it's not working, then it's not working. I'm a master of the obvious today...
I'd go with Nigel's recommendation. Start by making local duplicates of the projects and models with the UNC paths and then use the Monarch Utility to change those paths to one that works for those users.