The "context dependent filesystem" concept is a bit trippy, but I think it's a pretty neat solution to "some systems need a their own version of a file, other files ought to be universal".
The clusters had a shared OS image - that is a single, shared root filesystem for all members. To allow node-specific config files, there was a type of symbolic link called a “Context Dependent Symbolic Link” (CDSL). They were just like a normal symlink, but had a `{memb}` component in the target, which was resolved at runtime to the member ID of the current system. These would be used to resolve to a path under `/cluster/members/{memb}`, so each host could have its own version of a config file.
The single shared root filesystem made upgrades and patching of the OS extra fun. There was a multi-phase process where both old and new copies of files were present and hosts were rebooted one at a time, switching from the old to the new OS.
The am-utils "amd" known as its running process current use I don't have much to say as I've not much seen it as at least Linux distros have had autofs-tools quite long time. But -90 something am-utils was the thing we mostly used.
Adding: Oh, that made me remember we had then also user mode nfs daemon, which allowed re-exporting remote mounts, which was at times with smaller disks and always looking where to get it more if nothing but temporary storage great help. Current kernel based nfs doesn't support it any more.