Yesterday afternoon at 3:25 PM PDT Linden Lab posted to the Grid Status page that a previously planned service outage would be scheduled for this Wednesday. From 6:00 AM to 7:00 AM, so as you read this, the work is complete and an all clear was posted at 6:51AM.
During that time various backend services were offline and logins were offline. You can see the break in the graph of concurrent users.
Presumably this was the outage planned for last Thursday, before Tuesday’s problem when the login servers ran out of assignable user numbers. (I am guessing that fix is still in progress, but there is no official statement.) If this is the rescheduled outage then some hardware change was being made to some of the many components of the various databases.
While the Lindens think there may be some small improvement in performance they also suspect it will be small and probably not noticeable. I call these types of gains ‘stop watch’ gains, because they are smaller than we normally can notice unaided.