My blog (http://blog.stargrave.org/russian/) initially also used Git as a storage, where each commit is a post, and its log message is the content itself. If I add gitweb/cgit, when it will also feature web interface and Atom feed automatically. Later I added my own rendering engine to it: http://www.sgblog.stargrave.org/ for efficiency and automatic transparent gopher/gemini protocols support. Each post also can have some dynamic tags/topics attached, that are stored in git's notes branch. And later I also added ability to leave comments, by sending email message, that is committed in another git-note branch in the form of recfile (https://www.gnu.org/software/recutils/) plain-text comment. Everything is plain text, can be used with Git solely, but has an engine for better rendering and wider protocols support.
My 12+ years old website: http://www.stargrave.org/
Currently it is built as a Texinfo document. Previously I used reStructuted Text (Sphinx), Vimwiki, Gopher-compatible Perl scripts, FTP-viewable READMEs and static HTMLs. Contains much more data than even social networks will ask from people.
And again noone forces you to use multiple files too :-). You can literally have just single default.do file for the whole project. apenwarr/redo somewhere explicitly noted that. Separate .do files are only for convenience and ability to automatically depend on target's build rules independently from others, that Make just do not do at all.
Nobody forces you to use any kind of shell with redo. Its .do files can be anything you want: ELF object files, shell scripts, Perl, Python, whatever, just make it executable and obey trivial simple rules (stdout and three arguments).
FidoNet is still living, however it lacks at least strong authentication and data confidentiality (no crypto used at all). But it shows that people could be capable of creating global networks without unwanted third-parties.
For building store-and-forward networks I created NNCP several years ago and lack of connectivity, censorship (making no connection links) are one of the issues it aims to solve: http://www.nncpgo.org/Use-cases.html
Personally I used Python's tnote program, but, because of Python, it worked pretty slowly. So I rewrite on pure POSIX shell and it serves me for nearly ten years:
http://www.git.stargrave.org/?p=t.git;a=blob;f=t
Comparing to author's solution, it gives ability to briefly list notes, use multiple "namespaces", quickly add (without invoking the editor), delete or modify each specified note. Also there is no bashism and it works out of box on *BSD (that lacks bash) and GNU systems.
No, multiple output is not covered. gored passes on all my projects, even containing complicated redo-stamp/redo-always dependencies the same as apenwarr/redo. Not everything is passed in its testsuite, but I did not dig why. In practice it is complete replacement for my needs with the same behaviour.
I have posted link on HN to Dataforge UUCP because I have been interested in store-and-forward networks for a long years. All my email passing between notebook and my email servers used UUCP-over-SSH for years. But soon I create NNCP (http://www.nncpgo.org/) project and totally moved all my email/files store-and-forward networks to it. Possibly it could be interested to you.