THP,
I have to politely refute your first point.
Owning your own datacenter is no requirement to provide great service. In fact, the majority of hosting companies will flat out lie to you about this anyway!
Colocation (placing hardware you own in another companies datacenter) is an excellent option for hosting providers. You do not get the $1million + upfront cost of building your own facility, you get SLA's and 24/7 remote hands staff you can deal with directly and you have full access to your own hardware. You get all the benefits of running your own facility without the headache or need to bring in all the required network and datacenter engineers. It allows the hosting company to focus on what they do best, which should be building great hosting platforms and delivering fast, friendly support.
I colocate servers in 3 different locations, Vancouver, Montreal, Marietta GA. We currently have 16 racks full of servers and have our 17th coming online in the next few weeks in Marietta. I used to be in San Jose too, but closed that location down a few years back. The key to maintaining uptime and troubleshooting dead servers is to have your own hardware in stock. At each location we carry spare cpu's, drives, power supplies, raid cards, ram, etc. in stock.
Thanks to investment in devices such as APC remote powerstrips & KVM/IP, my staff and I can work on the server as if we were directly logged into it with a monitor & keyboard. When needed, we can remotely powercycle (reboot) the servers. Thanks to our contracts with our providers we can have remote hands support in our private cage within 15 minutes and have them open our servers, replace any needed hardware and get things re-racked.
I offer a 4 hour hardware replacement SLA. In 9 years of offering this SLA, I've never had to pay out a credit on it. Why? Proper planning, proper procedures, server grade hardware, proper hardware monitoring.
Be wary of providers who don't tell you EXACTLY what hardware you are getting. It's really easy to offer a "dedicated server" based on hardware picked up at an auction for pennies. Many big providers who own their own datacenter do this. It makes sense from a financial perspective, you can get 100 lot's of pc's and put them on a bakers rack for next to nothing. When a server dies, offer the client a new server from the 100 lot. Of course, this means the client will be down for hours, losing potentially all of their data but hey! churn is a part of the biz, right?
Well, no, churn should be pretty damn low and customer satisfaction should be priority #1. We lose, on average, .6 clients per month (yes, we actually track it). Of course, we have never had a month where we gained more then 38 clients (offering that in the interest of full disclosure).
I have had outages that were out of my control, such as a power main exploding in Vancouver, causing the generator to kick in and then the generator overheated and shut down. Took the whole building offline. Guess what? The company that owned the datacenter was offline too.
In that instance, we had to live up to our 99.9% uptime guarantee and credited quite a bit of money to our clients that month. Luckily our second biggest cost (bandwidth & colo fee's) were credited to us as well since our colo provider was down.
Another thing to remember... in the hosting world, EVERYONE is a reseller of something. It's the nature of the internet.
Oh and WHT is full of shills. Lot's of good info there, but be _really_ careful of accepting recommendations on there. I've seen more providers blow up on WHT... just do your research.... there are plenty of great hosts but it's very easy to appear to be a big company when in reality it's a kid in highschool....
Further full disclosure, I do not run a big company. Other then the people pictured on our team page, we employ 3 part time employees.