NFS was designed as a "state-less" file system.  This has several
advantages, but one of the disadvantages is that is doesn't handle
"state-ful" operations.  This type of removal requires the retention
of the knowledge that the original file system entry has been
removed, but the data must still be retained.  NFS handles this
by renaming the original file to the form ".nfs<something>" instead
of actually removing it.

It should go away when the process completes, but if it doesn't
for some reason, there are a couple of options.  On many/most
systems that serve NFS clients, there is a daemon job that sweeps
the file systems and deletes such files after a couple of days.
If you are sure that no programs are still active that might access
this file, it is safe for you to delete it yourself -- although
if you do this while some process still have the file open, you
will simply iterate on the creation of the strangely named files.
You probably also would be able to delete it from a program running
on the server system.

Just a little more detail if you're interested:  The use case to
consider here is that you have an NFS-mounted file open that you
have removed.  Now consider that the server is rebooted for whatever
reason.  Your client-side file system still needs to be able to
reopen the file, and it does so by using the ".nfs<something>"
form of name.

Here are some random references that Google came up witH;
        http://www.cs.indiana.edu/Facilities/FAQ/General/dotnfs.html
        http://www.seas.upenn.edu/cets/answers/nfs-files.html

 - dmw