I've been meaning to send this since last friday but been flat out.  We
figured out what was causing the processes getting stuck in "vnlock" and
the panic's and some other oddities we had noticed and the stability we
have had since then tends to confirm it.

When mountd is HUP'ed on a NetBSD NFS server it clears the in kernel exports
list and then reloads it so for a short period the server will send EPERM
responses to any NFS requests that happen while its reloading the table.
On the client side the processes getting stuck or the panics are presumably
caused when different operations get a failure from an NFS operation in a
way or of a type they are not expecting.

The reason we were seeing this is that for various reasons our filesystem
export lists change quite dynamically so mountd was getting hup'ed quite
often.

cheers
mark