Despite its better use of memory, it seems that "array too small" occurs in 8v0 also.  I used the following top file and a pdb file containing 70 residues and 2 calciums. After calculating the 5 loop models, it read in the models for clustering. Then it started to do the clustering, with the following log file (I've just included the part pertaining to the clustering. The optimizing of the loops went OK). As you can see, MBNDPF needed to be increased. I tried SET MAXRES = 5000, which isn't supported any more, and SET TOPOLOGY_MODEL = 1, which didn't change anything. What can be done?

top file
 INCLUDE                         # include the standard routines
#
 SET OUTPUT_CONTROL = 1 1 1 1 1
 SET LOOP_MD_LEVEL = 'refine_3' #'nothing''refine_1'(fast)'refine_2''refine_3''refine_4''refine_5(very slow);
 SET TOPOLOGY_MODEL = 1
 SET HETATM_IO = ON
 SET DYNAMIC_COULOMB = ON
#
 SET SEQUENCE = 'LobeX'             # the sequence code
 SET LOOP_MODEL = 'LobeX.pdb'       # the structure into which the loop is built
#
 SET LOOP_STARTING_MODEL = 1
 SET LOOP_ENDING_MODEL = 5 # how many loop models are calculated
 SET RAND_SEED = -34871     # to get different starting models each time
 SET ID2 = 0               
CALL ROUTINE = 'loop'                      # do the loop models
#
# Try to get an even better model by clustering (optional):
 SET MODEL = 'LobeX.BL00010000'     # one PDB structure
SET EXPAND_CONTROL = LOOP_STARTING_MODEL LOOP_ENDING_MODEL 0 0 0
 CALL ROUTINE = 'cluster', ID1 = LOOP_STARTING_MODEL, ID2 = LOOP_ENDING_MODEL
#
  SUBROUTINE ROUTINE = 'select_loop_atoms'
  # insertion:
    PICK_ATOMS SELECTION_SEGMENT = '65: ' '70: ', SELECTION_STATUS = 'initialize'
   RETURN
 END_SUBROUTINE

log file

fndatmi_285W> Number of residues <> number of atoms; atom code:       72       70  CA
fndatmi_285W> Number of residues <> number of atoms; atom code:       72       70  CA
fndatmi_285W> Number of residues <> number of atoms; atom code:       72       70  CA
fndatmi_285W> Number of residues <> number of atoms; atom code:       72       70  CA
fndatmi_285W> Number of residues <> number of atoms; atom code:       72       70  CA

Dynamically allocated memory at   amaxalignment [B,kB,MB]:      3288527    3211.452     3.136

Dynamically allocated memory at   amaxstructure [B,kB,MB]:      3315743    3238.030     3.162
mkfrw___326W> No selected atoms for frw position:       71
mkfrw___326W> No selected atoms for frw position:       72
malign3_328_> Initial framework positions:       72
mkfrw___326W> No selected atoms for frw position:       71
mkfrw___326W> No selected atoms for frw position:       72
mkfrw___326W> No selected atoms for frw position:       71
mkfrw___326W> No selected atoms for frw position:       72
mkfrw___326W> No selected atoms for frw position:       71
mkfrw___326W> No selected atoms for frw position:       72
mkfrw___326W> No selected atoms for frw position:       71
mkfrw___326W> No selected atoms for frw position:       72
malign3_330_> Framework, Cycles, RMS_frw(i-1,i):        1        4        0.0000

Least-squares dynamic programming alignment:
     FIT_ATOMS atoms for alignment     : CA
     Max dist from frw for equivalence :    6.0000
     Gap introduction penalty          :    0.0000
     Gap extension penalty             :    3.0000
     Numb of residues in framework     :        72
TOP_________>   522  649 SET ADD_SEQUENCE = on

TOP_________>   523  650 SEQUENCE_TO_ALI ATOM_FILES = ATOM_FILES 'cluster.opt', AL;
                      IGN_CODES = ALIGN_CODES 'cluster'


Dynamically allocated memory at   amaxstructure [B,kB,MB]:      3329259    3251.229     3.175
TOP_________>   524  651 TRANSFER_XYZ

transfe_505_> MODEL is an average of the largest cluster:        1.5000
openf5__224_> Open       14  UNKNOWN  SEQUENTIAL  nmemb.dat
openf5__224_> Open       16  UNKNOWN  SEQUENTIAL  occupancy.dat
trfxyz___> The largest cluster occupancy:
           Template -1, occupancy:    1    72
           Template -1, occupancy:    2    68
           Template -1, occupancy:    3    67
           Template -1, occupancy:    4    69
           Template -1, occupancy:    5    50
transfe_511_> Number of templates for coordinate transfer:        5
              After transfering coordinates of the equivalent template atoms,
              there are defined, undefined atoms in MODEL:      544        0
TOP_________>   525  652 WRITE_MODEL FILE = 'cluster.ini'

openf5__224_> Open       14  UNKNOWN  SEQUENTIAL  cluster.ini
wrpdb2__568_> Residues, atoms, selected atoms:       72      544      544
TOP_________>   526  653 READ_TOPOLOGY FILE = '$(LIB)/top_heav.lib'

openf5__224_> Open       11  OLD  SEQUENTIAL  ${MODINSTALL8v0}/modlib\/top_heav.lib
read_to_681_> topology.submodel read from topology file:        3
openf5__224_> Open       11  UNKNOWN  SEQUENTIAL  ${MODINSTALL8v0}/modlib/models.lib
TOP_________>   527  654 READ_PARAMETERS FILE = '$(LIB)/par.lib'

openf5__224_> Open       11  OLD  SEQUENTIAL  ${MODINSTALL8v0}/modlib\/par.lib
openf5__224_> Open       11  OLD  SEQUENTIAL  ${MODINSTALL8v0}/modlib\/par.lib
rdparf__223E> Internal error: array too small:  MBNDPF
              current maximum, current need:      250      251
recover____E> ERROR_STATUS >= STOP_ON_ERROR:        1       1

Dynamically allocated memory at          finish [B,kB,MB]:      3329259    3251.229     3.175
Starting time                                            : 2005/06/08  13:38:33.907
Closing time                                             : 2005/06/08  13:40:03.719
Total CPU time [seconds]                                 :      89.64