Earlier, I ran an experiment to see if we got what I considered to be true read consistency on a table. Initially, admittedly, I didn’t think it was possible without first locking the entire table. And I still believe that’s the case with updates I envisioned them initially.
I was, however, thrown for a loop when I discovered the SCN of the update that was being performed was after the SCN of an update that had been started after the update, which made me devise the theory that if an update reaches a row that has been updated since the start of the statement, the read consistency of the row is foiled and the query has to start back at the beginning to achieve true read-consistency.
Today, I’ll prove that theory either true or false.