A new index can be created or an existing one reused when a primary key constraint is created, which means a primary key constraint automatically attempts to create an index if one does not already exist. Moreover, an INVISIBLE index is indeed maintained by DML operations on the underlying table. This ensures it stays up to date even though it is not visible to the optimizer. However, a query filtering on an indexed column does not always mean the index will be used; the optimizer evaluates various factors before deciding to use the index. DROP INDEX does not necessarily prevent updates to the table, and a table column can be part of both a unique and a non-unique index.