Regression: sockets for 3.0.9pre1 ?

[grarpamp]

> There's no reason rsync would be doing anything differently from prior
> versions, so check what's wrong with your system.

Hmm, I have syslog *.*  going to disk, no kernel or
other issues reported.
3.0.8 was compiled on this system.
fsck -fy reports the source and dest are clean.

I do know that when these files were first laid down
in 2004-2005, that rsync back then didn't have
much in they way of handling all the metadata/links
regarding 'special' files correctly.

Only now, when making a complete new copy am I seeing
issues with rsync. In other words I had rsync -Haxi
--delete /1/ /2/ running often to pick up other source changes,
but those old deep dirs never had any src changes. Then /2
bit the dust, /3 is new, and new 'rsync mirror' needs made.

rsync has no problem with even deeper nesting of
directories and regular files, they all copied over just fine.

Since mknod is unaware of the actual mapping of
maj:min <-> filename, they're just numbers and names...
rsync should be reading and making the numbers and names
exactly the same... even if the numbers were remapped via
/dev/MAKEDEV since 2005. Doesn't matter since none of
this actually involves /dev.

The only other difference I noted with find in 2,000,000+
inodes was in block count with one sparse file, which is
normal given I did not use -S.

So I'm open to suggestions on what to look for.
But with zero system level errors, and find -ls
comparisons looking pretty good other than what was
noted... I'm not sure if there is anything else to check?