Everything else got paged out rather badly...
   I had vmstat -w1 running, at the end the numbers went:
   
    procs    memory      page                       disks         faults      cpu
    r b w    avm    fre  flt  re  pi po fr sr w0 c0 c1 f0   in   sy  cs us sy id
    0 1 0 170884    388  169 191   0  0 0 449 158 0  0  0  261  780 186  2  4 94
    0 1 0 170280    928  176 315   0  0 0 929 191 0  0  0  294 1011 229  2  2 96
    0 1 0 170012   1176  221 273   0  0 0 857 198 0  0  0  301 1000 239  1  7 92
    0 1 0 170524    624  101   0   1  0 0   0 122 0  0  0  224  691 153  0  4 96
    0 1 0 169900   1208   93 139   0  0 0 417 141 0  0  0  243  684 163  0  1 99
    0 1 0 170580    468  137   0   0  0 0   0 180 0  0  0  283  961 217  1  1 98
    0 0 0 116864  53928  178 267  15  3 0 835 169 0  0  0  271  913 231  1  5 94
    0 0 0 116864  53928   24   0   0  0 0   0 159 0  0  0  261  149  34  0  0 100
    0 0 0 116864  53928   12   0   0  0 0   0 148 0  0  0  250   97  19  0  1 99
    0 0 0 116864  53928   27   0   0  0 0   0 24  0  0  0  125  158  31  0  0 100
    0 0 0 116864  53928   12   0   0  0 0   0  0  0  0  0  101   89  17  0  0 100
    0 0 0 116864  53928   24   0   0  0 0   0  0  0  0  0  101  149  31  0  0 100
   
   Note that when the program exited ~54000 (~220Mb) of memory suddenly
   became free.  This can only have been the output file?  Which ought
   to be a prime candidate for being written out.
   
   Have I misinterpreted these statistics?


those are not pages those are KB.  filesystem cache is not reported
in these figures - you have about 114MB of ram used by non cache and
53MB ram free.  i imagine that most of the free ram (50+MB) came from
the mkisofs process itself exiting -- how large did *it* get?


.mrg.