Ah I see, sorry. What format is the row ID at the moment? The id would have to be a UUID for that to stand a chance of working, and not an integer like (eg) MySQL uses for auto_increment.
App Table row IDs look like this "[123456,987654321]"
the first number is the table ID and the second number is the actual row. If you want to use app_tables.my_table.get_by_id("[123456,987654321]") you still need to know what my_table is.
What I am after is something that doesn’t need to know what the table is ahead of time, it can still get the row. This is particularly useful if are storing linked rows as Row IDs in offline indexed_db or JSON objects. You can still get your rows on the server with minimal fuss.