That is partially what was done. Other than increasing the buffer size, he did a partition on the data. Together with the buffer size increase, if he created another table using the "archive" engine and kept pruning the data form the `domain` table to this archive table he would achieve the same thing with persistance.

All in all, it was a very nice solution and use case for the memory table! :)