The API is free, though applications are limited to a reasonable number of messages per month. That was put into place out of fear that a large company would use the app/service as a dirt-cheap platform to send large amounts of messages every day to its employees (like routing/dispatching directions, etc.) and burden the service. And rightly so, since shortly after launching, I got an e-mail from a large US company looking to use it in such a way (they ultimately decided not to use it). If such customers do come along, their API fees will cover them being on segmented servers.
Also, about your revenue model. How did you calculate that $3.99 would be sufficient to make profit AND give service to users on "all you can eat" terms? Wouldn't it be wiser to give one year for free (thats plenty of time to see whether I need and like your service or not) and then charge $9.99 per year after that? I mean, if this grows, I can easily see people eating up more bandwidth than this one-time $3.99 payment can bring into your pocket. If your TOS (sorry haven't checked) does not explicitly saying that in the future you may charge more, I think you may found yourself walking on a thin ice.
I went with $3.99 because Prowl is $2.99 and Notify My Android is $3.99. I'm not convinced that a yearly service is worth the overhead and I personally wouldn't pay to use a notification app that had a yearly fee.
The bandwidth and server overhead for running such a service are pretty low and iOS/Android push notifications are free, so I'm not really worried about monthly costs for normal users outweighing the revenue the app brings in. Right now Pushover is running on one of my servers that does other things, so it's not even really costing anything tangible. If costs start to go up, I can always increase the price of the apps to slow growth (or increase revenue if it doesn't slow growth).