I was really close to going with Heroku, I decided not to as I wanted full customisation options and I've been hosting rails apps for a long time and know what I'm doing.
If you're inexperienced in server administration, go with Heroku I'd say. It's a great service and works really well and if you do get a load of traffic, it's really easy to scale up. There's a guide here to using unicorn to get 4X the performance out of it:
You're correct in saying $40 per month for hosting until I gain more users. If someone would sign up to my "pro" plan I'd be breaking even :)
However I don't think that even if it does take off it will be as profitable as it sounds. It seems customer acquisition is going to be my biggest problem and so I'll probably end up spending lots on adwords and other promotional avenues.
I'm hosting it on linode, using 2 vps instances at the moment, one for the site and another isolated instance to run the code. The code running VPS is disposable so I can boot up new ones / restore from a snapshot if there are any problems with it.
With this setup, I've tested 200 concurrent users with no major problems, and would support easily tens of thousands of hits per day.
Having too many users is a problem I would love to have!
So far, I've spent about $40, all on hosting charges. As someone with design and development skills I was able to do all the work myself. I can't imagine how a non-developer could do anything similar without a huge budget.
Having said that, I haven't had much success. Although I've had a fair amount of traffic and a significant number of comments to the effect that my site is really well executed, I haven't had any sales so far two weeks after launch. I guess it's early days but it's hard not to be disheartened and question if the idea is viable. It's hard working alone and not having anyone to motivate and bounce ideas off of.