Why does performance degrade when many UPDATE, INSERT or DELETE statements are issued on a table that has an associated bitmap index?

Bitmap index blocks are not re-created on each DML operation; instead, the affected bitmap segments are updated in place, not dropped and rebuilt.

Oracle does not automatically rebuild a bitmap index after a DML operation; the index entries are updated incrementally, not rebuilt from scratch.

A bitmap segment in Oracle maps a range of rowids, so locking even a single row's bitmap entry requires locking the entire segment that covers that range. When multiple concurrent UPDATE, INSERT, or DELETE statements target rows whose rowids fall within the same bitmap segment, each statement must wait for the lock on that segment to be released, serializing DML and degrading performance significantly.

NULL values are stored in bitmap indexes and their removal is not a post-DML step; NULLs are simply represented as a separate bit position and do not require cleanup after each DML.