Gaussian98W memory requirements

Hello together,
 there are actually two questions. The first
 relates to an observation made by scanning the CCL
 archive. In almost regular time spans the question
 appears what to with the error codes given by
 Gaussian after a link dies. Surprisingly (at leat
 so me), a summuray of responses to these questions
 is apparently never posted. Does it mean that
 these error codes are only for the specialists
 > from Gaussian Inc., because no member of this list
 (THE experts, I suppose) has an answer to these
 questions so that no summary can be given?
 The second question relates to a hard (and
 unpleasant fact). A calculation is running on
 GAUSSIAN98W on a 3.4 GHz Pentium IV with 1 GByte
 working memory and 100 GByte disk space. Following
 the use of internal memory during a run it is
 found that the code never uses the internal memory
 to more than 20%. The calculation is a point
 optimisation of a lanthanide molecule with 8 water
 molecules (other ligand combinations were tested,
 the observation is the same - it seems to depend
 on the fact that a lanthanide ion is used). SDD
 pseudo cores with SDD basis sets is used for the
 lanthanide ion, the basis set for the light
 elements were successively reduced to 3-21G* (with
 no effect). The algorithm converges. Upon going
 into the next 'round', it dies with the following
  Compute integral first derivatives.
  Integral derivatives from FoFDir, PRISM(SPDF).
  Symmetry not used in FoFDir.
  MinBra= 0 MaxBra= 3 Meth= 1.
  IRaf=       0 NMat=   1 IRICut=       1 DoRegI=T
 DoRafI=F ISym2E= 0 JSym2E=0.
  Insufficient memory for a single shell
 combination in PrsMar called from PRISM:
  LenV=    57934341 MaxCom=           0
  Error termination via Lnk1e in
 Any suggestion to get around this nasty problem is
 highly welcome - as is contact to other CCL
 members doing Gaussian calculations with
 Best regards and thanks in advance
 				Passau, Freiberg & Toronto