Consider this sequence:

* I boot with root=LABEL=foo1 or root=UUID=[long]
* However I mistyped the UUID, or misapplied the label, or the root disk
  had to be replaced, or the rotofs is broken and I need to boot the
  rescue root
* I get dropped into the initramfs panic shell
* I mount the rootfs to /root manually

However, the current boot loop insists on finding the volume that
corresponds to the original root= argument. There appears to be no way to
override that.

What should happen: when /root is a mountpoint and /root/$INIT exists
(please check using "chroot test -e …" as /root/sbin/init might be an
absolute symlink), assume success.

Yes it should be possible to modify the kernel command line and reboot
instead, but sometimes it is not.