Without knowing a great amount of detail, it may pay to share the load across both controllers, assuming that both have public trunks. Given that the systems would be clustered and sharing all of the data it would spread the load a bit. Having the opposite controller listed as the fall back system, then the only time this problem would happen would be when one is offline, which you would assume is only for a short period of time.
Personally I haven't come across the 1000 limit, as there are generally multiple controllers spread across multiple sites, so the users register to their local site.