Cypress - adding retries to page object

Some more ideas:

On-the-fly retry changes

TLDR
Upside: simple
Downside: uses internal commands

There's an internal command that will allow you to change the retry count mid-test.

I stress the word internal as a warning that it may stop working at some new release.

With your code

export class Login {
  ...

  login(email, password) {

    const originalRetries = cy.state('runnable')._retries

    cy.state('runnable')._retries = 5    // WARNING - internal command
    
    this.username_input().type(email);
    ...

    cy.state('runnable')._retries = originalRetries  // restore configured retries value
  }
}

Retry by recursion

TLDR
Upside: good for flaky situations
Downside: explicit wait required

If you want the longer but more mainstream solution, use recursion.

You will need a non-failing check to handle the retry-or-finish logic, which will essentially mean changing cy.get('.profile-content').should('exist') to a jQuery equivalent and using an explicit wait (a downside for this pattern).

export class Login {
  ...

  loginAttempt(attempt = 0) {       
    
    if (attempt === 3) throw 'Unable to login after 3 attempts'

    this.login_button().click();
    
    // wait for login actions to change the page
    cy.wait(200).then(() => {  
      
      // test for success
      if (Cypress.$('.profile-content').length > 0) {
        return // found it, so finish
      }

      // retry
      loginAttempt(email, password, ++attempt)  
    })
  }

  login(email, password) {
    this.username_input().type(email);  
    this.password_input().type(password);
    loginAttempt()
    return this
  }
}

You can cut the cy.wait(200) down a bit and raise the maximum attempt, say cy.wait(20) and if (attempt === 300)

The downside is each retry uses up a bit more heap memory, so taking that too far will be detrimental - you'll need to experiment.

Set logged-in state directly

TLDR
Upside: By-passes flaky parts of test
Downside: Requires knowledge of login state

The other aspect you may consider is to cut out the flaky login altogether. To do that, you need to find out what the app considers to be logged-in. Is it a cookie, localstorage value, etc.

For each test that needs to be in logged-in state but is not actually testing the login process, set that state directly.

An example is given here recipes - logging-in__using-app-code

describe('logs in', () => {
  it('by using application service', () => {
    cy.log('user service login')

    // see https://on.cypress.io/wrap
    // cy.wrap(user promise) forces the test commands to wait until
    // the user promise resolves. We also don't want to log empty "wrap {}"
    // to the command log, since we already logged a good message right above
    cy.wrap(userService.login(Cypress.env('username'), Cypress.env('password')), {
      log: false,
    }).then((user) => {
    // the userService.login resolves with "user" object
    // and we can assert its values inside .then()

      // confirm general shape of the object
      expect(user).to.be.an('object')
      expect(user).to.have.keys([
        'firstName',
        'lastName',
        'username',
        'id',
        'token',
      ])

      // we don't know the token or id, but we know the expected names
      expect(user).to.contain({
        username: 'test',
        firstName: 'Test',
        lastName: 'User',
      })
    })

function login (username, password) {
  const requestOptions = {
    method: 'POST',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify({ username, password }),
  }

  return fetch(`${config.apiUrl}/users/authenticate`, requestOptions)
  .then(handleResponse)
  .then((user) => {
    // login successful if there's a jwt token in the response
    if (user.token) {
      // store user details and jwt token in local storage to keep user logged in between page refreshes
      localStorage.setItem('user', JSON.stringify(user))
    }

    return user
  })
}

(Editing the post to include the 2nd workaround...)

The entire idea of cypress is avoiding flaky test. If you have a login that sometimes fails and sometimes doesn't, then you have an unstable environment even before you start.

1. If all your tests depend on the login then it might not be a bad idea to include the retry at suite level (in the describe) but if only some of your tests rely on the login then you could group them all in only one suite and then add the retry at suite level.

Something like..

describe('Your suite name', {
  retries: {
    runMode: 2,
    openMode: 2,
  }
}, () => {
 
//number of retries will be applied per test
  it('logs in', () => {
    login.login(Cypress.env('userEmail'), Cypress.env('userPass'))
  })
  
  it('other test', () => {
})

})

2. Do the login through API and that one make it part of the tests that rely on the login (so it doesn't have to go through the UI and you won't get the issue). When testing the login itself then only there, do the retry for that test.

When you write:

Now our login page has the strange behavior of sometimes not responding.

You can rest assured, that your tests are not the problem.