You could look at CPU and memory usage of individual MaxDB processes and figure out what queries cause the big result set.
That's not really helping, as we need to check this over more than an hour, there are (hundreds of) thousands of statements handled in this time, and CPU/memUsage do not really correspond to the temp result immediately.
No, I really would need some kind of report which can be switched on and then evaluates the statements - like an SQL trace, but with the size of temp data result (or at least with the size of the ResultSet) saved for each query so that this can be evaluated afterwards.
Anyone else?