Detect empty (corrupted) files on automatic rebuild of the cache db

We had a node with a large number of job failures, like this:
http://bigpanda.cern.ch/job?pandaid=3106264518

http://aipanda053.cern.ch/media/filebrowser/017142a0-8b48-45b1-9cc3-fee11070992e/panda/tarball_PandaJob_3106264518_ANALY_VICTORIA/athena_stdout.txt

D3PD2MiniSL: error while loading shared libraries: /cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase/x86_64/root/5.34.19-x86_64-slc6-gcc4.7/lib/libXMLIO.so: file too short

I found that the file /cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase/x86_64/root/5.34.19-x86_64-slc6-gcc4.7/lib/libXMLIO.so is shown as having size 663740 bytes, but when trying to read it, it is like reading an empty file. No IO errors are reported. It is just empty.

[root@hermes-kvm007 ~]# ls -l /cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase/x86_64/root/5.34.19-x86_64-slc6-gcc4.7/lib/libXMLIO.so
-rwxr-xr-x 1 cvmfs cvmfs 663740 Jul 9 2014 /cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase/x86_64/root/5.34.19-x86_64-slc6-gcc4.7/lib/libXMLIO.so
[root@hermes-kvm007 ~]# md5sum /cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase/x86_64/root/5.34.19-x86_64-slc6-gcc4.7/lib/libXMLIO.so
d41d8cd98f00b204e9800998ecf8427e /cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase/x86_64/root/5.34.19-x86_64-slc6-gcc4.7/lib/libXMLIO.so
[root@hermes-kvm007 ~]# cp /cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase/x86_64/root/5.34.19-x86_64-slc6-gcc4.7/lib/libXMLIO.so /tmp/
[root@hermes-kvm007 ~]# md5sum /tmp/libXMLIO.so
d41d8cd98f00b204e9800998ecf8427e /tmp/libXMLIO.so
[root@hermes-kvm007 ~]# ls /tmp/libXMLIO.so -lh
-rwxr-xr-x 1 root root 0 Dec 2 12:29 /tmp/libXMLIO.so
[root@hermes-kvm007 ~]# touch /tmp/empty
[root@hermes-kvm007 ~]# md5sum /tmp/empty
d41d8cd98f00b204e9800998ecf8427e /tmp/empty
[root@hermes-kvm007 ~]# cat /cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase/x86_64/root/5.34.19-x86_64-slc6-gcc4.7/lib/libXMLIO.so
[root@hermes-kvm007 ~]#