Suddenly slow performance on CIFS/SMB. AFP OK.

I have 4x ReadyNAS units in use.  When copying files from one of them today (an NV+) the performance was woeful (would have taken a day to copy 6Gb or so).

I thought it was my machine's wifi connection at first, but copying from another NV+ it worked fine.

Both NV+ units were set up the same, so I looked at the network interface details.  The slow one was getting a  lot of TCP Retransmits (thousands every minute).

Not seeing any obvious reason for this I switched the switch port it was connected to with the fast NV+ to see if it was a duff switch port.  No difference.  Dog slow copying.  Lots of retransmits.  Also tried different cables.  No change.

I thought that perhaps the unit had got itself into a mess, so I restarted, checking the volume etc as it came back up. No difference.  

As a last ditch attempt I opened up AFP on the unit (my laptop's a Mac - the machine I was copying to).  The other ReadyNAS' I have do not have this protocol active but I thought I'd check it.  I then tried the copy again, connected via AFP rather than CIFS/SMB, and it was back to full speed.  No TCP Retransmits.

I haven't switched it back yet as I want to keep the performance going...but does anyone have any ideas what could be causing this?  Or what other checks I should do to trace it?  It seems to be software related if a different protocol works fine over the same physical hardware...?