When using InnoDB engine for temporary tables, reclaiming of free table space does not seem to work.
Periodically performing statements that force the server to write into InnoDB temporary tablespace eventually results in the reaching of maximum size of temporary table space or even a full disk.
I was able to reproduce this behaviour by execution of following statements more times even when the connection between the executions is dropped:
After each execution grow the file /var/lib/mysql/ibtmp1 in size
The file /var/lib/mysql/ibtmp1 should grow in size only after first execution. After second an further execution should the file size stay eventually unchnaged as the space claimed by the first execution should be reused by the further executions.
Last version where the expected behaviour was observed: 10.5.11
Workarround: explicitly setting engine for temporary tables to MyISAM