Rewards seem nice, yet not everyone who wants to play
can pay, or the math overhead is crushing, or it becomes
centralized. Definitely worth trying, especially if it fits some
usage model.

Another form is to just let the network use whatever
CPU, RAM, DISK, NET that you're not currently
using, or give it whatever limits you want. In short,
set it and forget it. Let the network figure out how
to best use your node to support the network.
Maybe it's a strictly filesharing network,
or a general purpose network.
That's on the "Hey I just want to donate
this because it's cool like Seti@Home, etc.

Users actual use of the network would
be through different apps... be it submitting
infohashes, or compute jobs, etc.

Does eliminating all the reward tracking overhead
provide substantial resources back to support
free use.

ie: Most people and their computer resources sit idle,
probably more than enough to provide back whatever
multimedia they want to consume.
If true, all balances out, no need to bother track accounting
with "pay to play" style system?

I like "pay to play" as it offers at least some
firm guarantee to the consumer offeror.

But an accounting free system is more fun as in free beer :)

Hybrids might work too.

https://en.wikipedia.org/wiki/Exabyte
https://en.wikipedia.org/wiki/Zettabyte
https://en.wikipedia.org/wiki/Yottabyte
https://en.wikipedia.org/wiki/Orders_of_magnitude_(data)

100M users donating 10GiB slack space is about 0.93 EiB
of non redundant storage, excluding overhead.

Example, at 4x redundancy, that probably easily covers
lossless versions of all movies (at least 1080p)
and all audio (FLAC), all wikipedia, all OS and apps.

Approaching mini-NSA scale... not a bad start.