Definitely, I can see your preferences on pay per use. You have some good points on how pay per use can bring great benefits for the user. Specially, I can see how attractive would be this option for the infrequent users who would be willing to pay for a use rather than a subscription which obviously wouldn’t fit them at all. I had experienced this problem myself many times, for example Adobe Photoshop Lightroom, it’s in my opinion the best photo editing App for IOS, with a ridiculously expensive subscription. That makes completely sense for professional photographers. However, they are missing all the amateurs.
On the other hand, I can see why this income model has not been as spread as it could. The main reason, it’s that infrequent users who are willing to pay for the subscription are giving a greater benefit to the company. The most typical scenario, are the gyms, where they have tons of users who are paying but not using the service. I think it would be a great research which of this model returns more profit to the company. (More users vs more profit per user, after all)
An disadvantage of pay per use for the user is that it’s very hard to predict the total cost of the service. This might be scary for the vast majority of users. The target of AWS is not the same that Spotify. Who is more willing to make some maths? In order to make this model more user-friendly, it would be necessary to present it in a very clear way. I would love to see some propositions :)