We're already down to 99.7%.
If we do some quick math... This currently translates to outages of:
2h 11m 29s
6h 34m 27s
1d 2h 17m 50s
There's a margin of error, as the uptime monitor only checks once a minute. (You don't want it constantly pinging the server.) When it detects an error, I think it then checks every 30 seconds to see if the site has resumed responding.
So, my more-or-less educated guess is there's something at the server side that's dropping the connections. As this is a leased server/VPS (I believe, I know it's not being run out of @Rob
's basement), there's not a whole lot our Dear Leader Admin can do about it other than yell at his hosting company, perhaps using the uptime monitor to show that they're not hitting their SLA (assuming he has a contract that stipulates a minimum amount of downtime, which many do).