Retry re-runs the deployment instead of creating a second one

When you change your password on your profile, or reset it from the sign-in page, the panel no longer accepts the password you are already using. You are asked to choose a different one and your password is left as it was.

Environments on deployment pipeline version 1 turned away a release archive compressed as .tar.zst as soon as it was uploaded. They now accept it, from the Deploy page and the deployments API, as version 2 environments already did.

A single deployment step, such as a long database migration, was stopped after 2 hours and the deployment failed. Each step can now run for up to 8 hours, so the deployment carries on to completion.

When a site's containers were replaced during a deployment, the next step could connect to the container that was being shut down, and its work was cut short when that container stopped. Deployments now wait for the old container to finish shutting down first.

The list of deployments on an environment, and the full deployments list, no longer need scrolling sideways to reach a deployment's status, dates and buttons, whichever way you opened the page. On smaller screens the Tasks and Duration columns step aside to make room, and both are still on the deployment's own page.

A deployment stopped part way through could not be retried: it sat as queued and never started on its own. Pressing Retry once now starts the replacement straight away, and where a deployment does have to wait, the page says what it is waiting for.