Using the following example pipeline

stage('stage_1') {
  parallel (
    's_1_1': {
      node() {
        wrap([$class: 'Xvfb']) {
            sh 'sleep 10'
        }
      }
    },
    's_1_2': {
      node() {
        wrap([$class: 'Xvfb']) {
            sh 'sleep 10'
        }
      }
    }
  )
}

the Xvfb binary is always invoked with ":0" as screen name, even if the documentation suggests the screen name should match the executor number + offset.
Manually specifying displayNameOffset has no effect.

The only way to make it work in parallel is by setting autoDisplayName, but we've experienced random failures with that option too. It would be great for the default behavior to allow parallel jobs without any manual configuration.