Okay, finally got to the bottom of the weird scp stalls.
Turned out to be
a weird OpenSSL quirk brought on by the truly terrible kernel & glibc versions we're using...
Tweaked a few shell related things (most prominently, getting rid of the fugly blank hostname in a few places, without actually modifying said hostname), and cleaned up up the sshd/telnet startup script, fixing a nasty issue with broken pidfiles those last few days (which would allow spawning multiple instances of inetd, which wasn't great, but mostly harmless, and could only ever happen if you hot restarted nickel).