I get 1000 objects in my XML as desired, however the proxy memory usage goes crazy. There are 600,000 objects in the trade cache, the paging (which I assume turns into a limit filter?) seems to bring all the results into the proxy an then limits the results. This is different to the native limit filter, which seems to filter at the cache node level.
curl http://lons00xxxx:5050/Trade.xml;start=0;count=1000; > c:\trades.xml