There's been a couple of attempts at getting SCL into fedora guidelines which haven't gone anywhere (last I read IIRC was that what might be considered valid for RHEL as a corporate entity is not necessarily valid for Fedora so no SCL as FPC don't agree with it).
This actively hurt EL6 as some projects, for example owncloud, had to be dropped and others, such as nextcloud or letsencrypt/certbot, could never be introduced.
The EOL date for EL7 is still a long way off but anything relying on PHP, for example, will soon need to be removed from EPEL7 as projects drop support for the long EOL'd php5.4
I'd really like to see an effort within EPSCO to address this. As Red Hat relies more on SCL for recent versions of certain applications there's less chance of a rebase of these which will affect what we can actually package in EPEL.
This could potentially tie into your point 3 as CentOS already had the SCL SIG that builds on the Red Hat SCL packages as an upstream.