How temporal could address the issue that newer version workflow get replay on older code due to the fact that both old and new instance coexist during canary deploying?
In most cases, the replay of the workflow that made progress using the new version on the old worker will fail. It will be retried and eventually hit the updated worker to continue successfully.