I don't think it's quite a simply answer like X% uptime.
A lot of the decision is the 'quality' of the downtime. For example:
* Is it generally stable, but only going down in peak load situations (obviously no good if you know peak load occurs when you are doing critical new product release/sales etc, but may be okay if you peak load corresponds to non-critical events).
* How long is the outages? If it goes down for 1 minute each day, that may be more acceptable compared to '100%' uptime except one day when it goes out for 365 minutes.
All of that affects the impact that the downtime has on your venture - it's quite possible you might have 'poor' uptime stats, but it has zero effect on your business, in which case the cost/effort savings from using a 'poorer uptime' hosted service may be worth it.
I think it's more important to know _when_ the downtime occurs vs. the percentage itself.
If the site is down from 12am to 6am due to some botched maintenance, I really don't care, as traffic will be minuscule, if not non-existent for our core market. And 6 hours of downtime is a killer for uptime percentage.
The "Jekyll is so easy! It's the easiest thing I've ever maintained!" argument reminds me of Andy Rubin's now infamous quote about the openness of Android: "the definition of open: "mkdir android ; cd android ; repo init -u it://android.git.kernel.org/platform/manifest.git ; repo sync ; make"
The point of Tumblr is I don't need to maintain _anything_!
