There are no traces on the client in /tmp from the actual run but the
  amcheck logs are all OK.

Assuming you mean amcheck debug files in /tmp.

I would run 'tcpdump -s1500 -wamanda port 10080' while doing amcheck
and amdump, and see if amandad is started during the amdump.

I would say it's possible one of amcheck/amdump is messed up - they
should be doing similar rpcs, but given the client upgrade and
specific client issue, that doesn't seem too likely.

It could be that amdump is invoking some unhandled option in the
control packets that amcheck isn't, but this seems unlikely too.

-- 
        Greg Troxel <gdt AT ir.bbn DOT com>