I would check the local file caches on the application services (specifically in any resource directories). If the boxes are the same and were not wiped completely clean, that could be the cause (not familiar with the exact processes of cleaning/refreshing).
Lastly, if there are any CDN caches that might be upstream from the Jive instance, using a CDN flush or header override to do a pull of the resource would probably help. (this is more likely in hind sight)
Check your routing, if you are getting routed through CDN for the request, then take appropriate action to disable/refresh accordingly. hope that helps. =)
We have a case open but it's drawing a blank. I've run your suggestion past the engineers and I'll let you know how it goes.