How does Cypress retry-ability work?

Retry-ability is Cypress's answer to asynchronous UI. When a chain ends in an assertion, Cypress reruns the query that produced the subject and checks the assertion again. It keeps doing that until the assertion passes or the command times out. This is why a test can assert on a value that appears a moment later without a sleep.

cy.get("[data-cy=status]")
  .should("have.text", "Ready");

Only the last command retries

This is the detail most people miss. If a chain is cy.get, find, first, should, the retry applies to the tail of the chain, not to every step independently. Cypress restarts the query from the last command that can be retried, which is typically the query immediately before the assertion. A stale subject earlier in the chain is not magically refreshed.

cy.get("[data-cy=row]")
  .find("[data-cy=status]")
  .should("have.text", "Paid");

• Retry applies to queries and assertions at the end of the chain
• A .then callback is not retried as a whole
• Network-dependent changes need an intercept or a fresh query, not a longer sleep
• Timeouts bound the retry loop; a longer timeout is not a fix for a real race
• Actions such as click are not retried blindly, so assert the resulting state