I don't think this wonderment really applies to non-dedicated hosting (however awesome Slicehost might be). When dealing with shared hosting of any sort (VPS included), it's a reasonable assumption that the host will stay fully on top of hardware issues, since you are not paying for a physical box.
If something craps out in one of my dedicated machines, I'll alert the host or NOC to have the fault looked into and fixed. If something craps out in a box hosting my home directory or VPS or what have you along with numerous others, I already expect that to be automatically detected and fixed within an acceptable timeframe. Ideally, in a VPS environment, my machine(s) should be migrated to a working box in the interim.
From my perspective, this is the only way that one can do business now-a-days. It's not because people demand good customer service or that they have a responsibility to do it. It's because not doing it takes more work and more effort over the long run.
So, Slicehost has something monitoring their boxen and alerting them and they deal with the problems. What if they didn't? They'd be getting emails from customers, have to go check it out, most of the time it would be the customer's fault (they powered it down or turned off the Apache service or set the firewall rules wrong) and they'd waste all this time and effort for nothing. It wouldn't make their service better. It's just spinning one's wheels and worse, it doesn't scale. If you're constantly checking boxen, you're going to have to hire more staff as you grow to also check on boxen.
And this is something that one can automate. It both lowers the amount of labor, effort, confusion, and swearing involved and raises the level of service offered.
I see it the way I see my code tests. Sure, they're annoying to write sometimes, but what's more annoying in the long run is having random errors that I don't know about or worrying that a change I made broke something not readily apparent. My tests make my software glamorous because, while they aren't glamorous or cool or anything, they prevent me from having to do even less glamorous things later as I scramble to fix something.
I just couldn't imagine working where I didn't have things watching my back like that. It would be a nightmare of "oh $#*^! what's wrong" and lots of time/energy/effort wasted on something even more annoying than being prepared.
I don't care what you THINK you're selling, you're not. You're selling customer service.
Lets look at my favorite bar. I go there ALL the time. It isn't even remotely close to my house (anymore), and the beer is silly expensive (because it's really good). OF COURSE I could run to the local liquor store and buy it, OF COURSE that would be cheaper, OF COURSE I could then enjoy as much of it as I want in the comfort of my living room without having to worry about driving home (meaning I usually only get to have 1 or maybe 2 beers).
So why do I keep going? Because all of the bartenders know me by name, they all know what sorts of beer I like (they have over 500 beers here, so them knowing what I want helps if they're making suggestions) and my glass is NEVER empty (If I'm having a friend drive).
On multiple occasions, the bartenders have said "Hey, I just put 20 free credits in the jukebox, you should go over and pick some songs".
One time, I had ridden my bike there and it started raining...I was about to head home when the bartender set a beer down in front of me and said "Why don't you wait it out, this one's on me".
It's the exact same thing with hosting, or bananas, or toilet seats, car alarms, computers, lightbulbs, EVERYTHING.
OF COURSE I could set up a box on my own, OF COURSE I could find a bandwidth provider and pay them, OF COURSE I could round up enough of my friends to make the cost of running the server less than what I spend a day on coffee, OF COURSE I could keep spare hardware on hand for when it goes down, but you know what?
I don't! I don't do it because to ME, it is worth buying the customer service from slicehost to have them do it for me. THAT is what I'm buying. I'm not buying server space, or bandwidth, or anything like that, I'm buying SERVICE.
Some companies are have figured this out (the one I work for is one of them), and those that have are thriving right now.
(if anyone here is curious, the bar I'm talking about is called 'Papago Brewery' and it is in Phoenix, AZ...if you're in Phoenix, stopping here is worth it, trust me).
I don't think that a hosting company that monitors their servers deserves this much credit. I have a few co-located boxes for personal projects and, even though the services they run aren't that critical, I still monitor them using Nagios. It's not hard and it should be something that every systems administrator sets up as a matter of course.
I do, however, agree that 20 minutes down-time for a hardware failure is impressive. I wonder if they just yanked the disks from one box and jammed them in a spare?
Slicehost is awesome. I have received a few such emails before, mostly alerting me that a slice needed to be rebooted, etc.
I've tried Joyent and had a HORRIBLE experience, and I've had a frustrating experience dealing with EngineYard's overly complicated sales process.
I was worried that the Rackspace acquisition of slicehost would kill the awesome customer service, but it seems only to have made more resources (for things like bigger slices) available.
I couldn't recommend any company more highly than Slicehost.
I had a similar experiance when I posted on twitter about my interest in switching to Ec2 from slicehost. About 5min later I got a reply via twitter from slicehost asking if there was anything they can do to improve the service.
I've had very good experience with Rackspace support. They managed to bail me out in a matter of minutes when I fat-fingered an iptables config and made the box totally unreachable.
Paid advertisement? I've been to a number of VPS hosters that do just that and much more - ServInt, Knownhost, liquidweb etc - and they have better plans too.
The 64bit only stuff really hurts them, though. In my comparison (Google for it, I feel bad for linking to it too much), Slicehost is a bit more expensive than Linode to start with, but it's not a huge difference; it's something you could ignore. Throw in the 64bit vs 32bit though, and you really see a big jump in price that's difficult to justify.
If something craps out in one of my dedicated machines, I'll alert the host or NOC to have the fault looked into and fixed. If something craps out in a box hosting my home directory or VPS or what have you along with numerous others, I already expect that to be automatically detected and fixed within an acceptable timeframe. Ideally, in a VPS environment, my machine(s) should be migrated to a working box in the interim.