Follow related operations are not associated with a dedicated transaction, they are stored in the blockchain using
custom_json transactions (see TxCustoms).
Storing the details of a "follow" operation in a JSON string is pretty annoying because this requires deserializing the content of the JSON string before you can analyze its contents. Even if the SQL server has JSON functions to perform such operation, this will slow down queries as the server has to process every row before being able to filter data. Unfortunately, it is not possible to create an index on values contained in the JSON string.
Searching for followers or reblogs is all but efficient, especially when you know that the TxCustoms table contains several million rows!
Moreover, the format of the JSON string of "follow" operations has changed over time. This has to be taken into account when parsing the data. Writing an efficient query can quickly become cumbersome.
The table is a non-transactional table that enables one to know who is following who. It does not store transactions, meaning it does not expose historical data and does not allow one to know when an account started or stopped to follow another account.
the name of the follower
the name of the followed user