I am not sure what oldest snapshot means, but these two are the
general indicator of database health.

> Oldest transaction 9181346
> Oldest active 9181347

In this case it looks good to me, the gap is < 20000.

Perhaps someone else could shed some light on your issue, and why a
restart may help. Is it possiblly a resource thing? As in is the CPU
pegged at 100% due to some intensive operation? Restarting the server
will obviously stop this operation which may explain why performance
comes back.

Adam