It may make sense to instead run two Coder deployments, accessible under different URLs:

• The latest release
• The latest nightly

The scaletest template can then be run on each independently.

@johnstcn that's definitely an option. In that scenario I would still like for us to have the ability to:

1. Prevent automatic updates (while running a scaletest or testing an older version)
2. Ability to easily change the deployed version (ideally accessible in an UI, like the runner)

We don't have to support 2 immediately, but we should keep it in mind. Why do we want it? For instance, say we detect a regression but are not sure when it started, we could then do a bisect type of flow.

That said, the original proposal has two pros:

1. Everything is in one place (version, running)
2. Enables the ability to test multiple versions either sequentially or concurrently

For now though, I'd classify these s as nice-to-haves.

The Coder deployment on big.cdr.dev is now managed by FluxCD with configuration hosted on https://github.com/coder/big.cdr.dev. To override the version installed, we can either edit the configuration or suspend/unsuspend the HelmRelease directly.