Unluckily I have no experience with combination of DBI and Oracle, specially with clobs.

But, naturally, I do not know the expected size of your 5000 clobs. If, for instance, 1 clob is maximally 64kB and 350MB of memory is acceptable for your Perl process, let you pull 5000 rows at once... When I implemented data replication between MySQL, Postgres and MSSQL servers, I have chosen 1000 maxrows.

I use the result mostly in this manner:

my $aref = $sth->fetchall_arrayref();
foreach (@$aref) {
    #now $_ is ref to an array holding one result row
    for (my $i = 0; $i < @$_; $i++) {
        print "$i: $_->[$i]\t";
    }
    print "\n";
}
[download]

If $aref in your question is the return of fetchall_arrayref, it is reference to an array, which items are references to an array too. So,

$$aref[0] or $aref->[0] is the first item of the "upper" array, i.e. reference to the array containing first data row,

$$aref[$rowindex][$colindex] or $aref->[$rowindex]->[$colindex] is the $colindex'th item of $rowindex'th data row.