There are systemd-free distros of Linux, you know. I can pretty confidently state that it will remain that way unless systemd should start to integrate itself into the kernel.
Well, yes... Most importantly RHEL6 / CentOS6. Those of us using Linux in business/enterprise settings are mostly running that, and that's mostly what we care about. The time limit on that is what we're sweating.
RedHat (Inc.) seems to be undervaluing its Good Will in terms of building an enterprise platform that goes well beyond RHEL subscriptions. EL users don't care about most of the systemd "feature" set (with the possible exception of easy(-ier) cgroup management), since most of the rest either doesn't apply or attempts to re-solve and already mostly-solved problem (eg, service monitoring and restart scripts). The cost is using less mature, less modular, less tested code with more common failure points, which might cover 80% of your needs but makes the other 20% of system customization really, really difficult, because apparently shell scripting is a Sin now.
Oh, and most of your config management that worked pretty similarly between EL5 and EL6 has a *lot* more of a delta to work with EL7.
"Forking Fedora" doesn't seem like it will happen, even though there are fewer and fewer non Kool-Aid drinkers there who think keeping your options open is a good thing.
Do you know what I'd like for EL8? Fork EL6, update all the non-daemon RPM versions to their current Fedora level, and run systemd as Just Another Daemon, akin to xinetd, supervise, or your cluster management software.
We get more reliable and more deterministic startup and shutdown process using the previous initscripts toolset and regular /sbin/init, and those who want the management capabilities of systemd for services can still use it, albeit with it not functioning as PID 1. I'd pay for that.