...
- Read the next datablock from the disk for the main table (blocks read during table scans are not pinned to the shared cache for large tables)
- Perform an index scan for a TOASTed counterpart to find rows that relate to the row in the main table (index blocks can be either in the shared cache or not)
- Retrieve all those rows that contain a JSONB value (in most cases db blocks must be read from disk)
- Decompress the value of needed and apply a function or make a comparison to the value
...