1script - 5:16 pm on Dec 24, 2012 (gmt 0)
Thanks for your input, guys! Yes, I do understand the dangers of downtime and have some horror stories of my own to tell (specifically @backdraft7: took 6 months to recover from a 72 hrs downtime-induced penalty/deindexing/whatever you want to call it) .
I'm sure there will be some downtime, especially considering that the hosting company prefers to use Russian and Indian techs for any work like that and I'm probably not going to be available for some time after they're done because of the time difference. Anyway, I'll keep my eye on it during and right after the move, as soon as I get access to the new server.
Regarding DNS propagation: did anyone try to lower TTL to something like 30 seconds (or less) a day/two in advance to speed up the switch? Although Google seems to be picking up DNS changes in a matter of seconds anyhow - I often see them pounding on a brand new IP (using proper URLs) in a matter of seconds after an IP change even when TTL is at default 1800 (30 minutes). I'm not even sure how that's possible - it looks like they don't actually completely trust DNS records anyhow and re-load them every time. Also possible that every individual instance of the bot has to get a fresh DNS record each time it first gets to a site, despite the previous instance of the bot having requested the DNS less than TTL time ago.
But as far as the actual fact of changing IPs (downtime dangers noted) - it seems that the consensus is: it's a pretty benign task.