Was just checking out blitz.io, which looks really cool (and fun), but I was stopped pretty early on by the pricing. I don't understand why the length of a rush is tied to the number of concurrent users (for pricing purposes). 1-minute tests are basically useless, other than checking if the site's going to just immediately fall over at a given load. I don't even trust 10-minute tests. To adequately determine the capacity of a system, I'd want to run a test for at least half an hour and ideally an hour, as the various components of the system go through their various operations (GC, for instance). I don't always need 50,000 concurrent users, but I do want the ability to choose longer rushes. Do you have a mechanism for that? If not, could I suggest you add one?
Less than half an hour later (nice), Michael Smith from blitz wrote back to tell me new pricing was on the way.
Well, it's here, and it makes so much more sense: You buy credits (they start at $1 USD = 1 credit), and one credit is worth 1 minute of 1,000 concurrent users. No longer is the length of a rush tied to the number of concurrent users, you control those variables completely independently. A 20-minute rush of 1,000 concurrent users costs 20 credits. 20 minutes of 10,000 concurrent users, 200 credits. And instead of being clock-on-wall time for a set period (which apparently it used to be), you're only charged for the time the rush is actually running; if something goes wrong you can stop it early to avoid spending credits unnecessarily.
As part of this, they did discontinue their old free level in favor of topping up all accounts to 10 credits each month (even accounts that haven't bought credits). There's some discussion about what that means for their continuous integration clients, which is probably worth a read.
But fundamentally, well done blitz for listening to your customers (clearly it wasn't my note that got them working on a new model, that was just happy timing for me). Now I can take a closer look.