I think the key to making it work is providing easy to understand controls over how the spending is done, for our project, at least, it is a fee for service that is paid a little in advance of use, so it skates a line something between a subscription and a pay-per-use (paywall). You just set a budget and the client maintains a pool of sessions it uses to perform messaging tasks.

In Indranet there is a risk of non-performing relays so in part of the remedy for this there is a feedback system that affects the ordering and frequency of advertisments of relays.

Another thing that helps in this kind of pay for use micropayment scheme is history of usage so the user gets some idea of what a reasonable budget is as they get more familiar with using it. There will also be filters to enable one-time and always on preferences for specific services or end destinations for tunneling out, geoip or social network based.

Lastly, because it is a peer to peer system, it is possible to also link your clients and relays that you run, together, so it becomes self funding, potentially. This is something that would be impossible with conventional payment schemes, without a centralised server managing much of it.