How often do you really need to do a full crawl? Usually an incremental
will do. Beyond that, RAM is cheap; do you have room to upgrade the box?

1500 pages  is nothing. This would mean very little effort on the crawler. We
have one doclib (!!) with 350.000 documents in it. Full crawl takes about
1,5 hour. So again 1500 pages would not mean anything. Off course we are on
64bit 4 quad core machines. 1 machine does indexing an querying

So perhaps you can limit nr of simultaneous documents for you content
source. How long does your crawler run before it reports the error? Maybe it
has problems with certain pages? I have noticed bad error handling in the
crawler if it encounters problems with for instance a "strange" element in a
custom site  def.

Off course you can split the pages over several libraries, but i doubt this
will help  you because again this volume should not be any problem.