Keeping PostgreSQL on AWS and shifting traffic a few percent at a time behind the same public URL is a realistic way to structure a migration, since it avoids moving the hardest piece on day one. Putting 'Prove a Retry Never Charges Twice' in Part 1, before any cloud work, is a nice touch for a payments app. Recording the AWS baseline before planning the move also gives the later load tests something concrete to compare against.