I think the Berkeley guys had an underground
pipeline to Bell labs and some stuff got out that way. :-)
It was not underground at all. Tools packaged in BSD came from all over the community. style and diction were released into the wild by themselves before the were packaged into an AT&T USG UNIX or Research UNIX release. It got them personally directly and had them installed at Tektronix soon after first publishing and a talk about them at USENIX (IIRC that was the Boulder conference in the "Black Hole" movie theatre.
Since I had a minor stake in it (as my first C program) fsck is another good example of the path to UCB . Ted started the predecessor program when he was at UMich (with Bill Joy). He did his OYOC year and later a full PhD at CMU. He was one of my lab partners in his OYOC year. fsck was a we know it now was done during that time ( and I helped him a bit). He was bring the sources back and forth from Summit to CMU (at the time in an RK05 or sometimes a bootable DOS tape image of one - I may still have one of these). I believe he gave a copy of the sources very early to wnj -- which is how it ended up in 4.1BSD. I don't think it was in the original 3.0 or 4.0 packages as it was not in V5, V6 or V7 either. I believe it was released in PWB 2.0 - not sure and Minnie does not seem to have them.
I'm pretty the SCCS and cpio sources came through one of the PWB releases (1 or 2) that UCB got from AT&T.