Two things were done: first, actual Deadlock errors are not caught and retried; second, there's a 0.25s delay between retries to let one or the other contentious process to move on. The customer reports that the changes have allowed the tool to run without dying on a very large, long-running table.
Two things were done: first, actual Deadlock errors are not caught and retried; second, there's a 0.25s delay between retries to let one or the other contentious process to move on. The customer reports that the changes have allowed the tool to run without dying on a very large, long-running table.