This is a good point — how would you change their pricing model then? The difficulty of supporting creators will grow with users, would a tiered model that starts charging earlier (but charges less) be better?
Maybe it's better to just do a single no-bullshit price and run unlimited ghost instances for people -- and once you know the scale that a particular instance can take, different tiers for the very large instances. Maybe one price (let's say $10) + custom? But at this point you'll be overwhelmed by users with very large followings which generate tons of traffic.
And if you go by a strictly bandwidth model, then people are vulnerable to overages...
Absent native phone apps, the only way to reliably send notifications to people is SMS and email. This is the world that was wrought when we let ourselves be duped into handing over the whole internet to two cell phone dealers (Google and Apple), and letting them tax real time communication worldwide via their app stores. And the cost of sending email is not just the AWS SES cost per 10,000 mails but the cost to render the template for each email, too. It's not hard to get into a situation where one server relaying mail for just a handful of domains can't keep up with its own queue in any given day.
I think if you're trying to cater to paid tier users, segmenting users by free/paid and only serving free tier users the cache without any sort of interactivity features is the first thing (no comment posting, no upvoting / downvoting, etc).
Finally, yes, I think the scaling cost has to be passed on to the site owner, I don't think you can shoehorn a very elastic service in terms of hardware / network demand into a fixed monthly price, without subsidizing successful users on the backs of those previously mentioned hobbyists.
Lastly, on a more philosophical level, I think media producers / publishers need to get together and stop nipping at the heels of Facebook, Twitter, Google, Amazon, and Apple like puppies following their masters around. I mean, the whole newspaper industry let itself be duped into poor staffing decisions by Facebook lies about video metrics. Which is funny considering Facebook's whole business model is taking other media businesses' ad revenue and keeping it for themselves. I guess media execs never heard the one about grandma and the big bad wolf in their MBA programs...
Maybe it's better to just do a single no-bullshit price and run unlimited ghost instances for people -- and once you know the scale that a particular instance can take, different tiers for the very large instances. Maybe one price (let's say $10) + custom? But at this point you'll be overwhelmed by users with very large followings which generate tons of traffic.
And if you go by a strictly bandwidth model, then people are vulnerable to overages...