The problem:
Deleting segments in batch triggers reorganize of the memory, which takes a lot of time for big memories.

The solution:
If amount of segments in batch deletion query is less than 300 do not call entriesdelete api endpoint in t5memory, but instead delete segments one-by-one in a loop. Segments should be deleted in reverse ordering (from the latest to the first to avoid btree rebuilding error in between).

In case TM btree becomes corrupt while deleting, memory needs to be reorganized and then deletion continued.