Certain unique jobs states configuration can block all jobs from retrying

[freemer]

Environment

• Oban Version: 2.19.2
• PostgreSQL Version: 16.6
• Elixir & Erlang/OTP Versions: Elixir 1.17.2 (compiled with Erlang/OTP 27)

Current Behavior

Having jobs with a state of retryable that were enqueued with unique: [:available, :scheduled, :executing] can cause all jobs with a state of retryable to not retry.

Expected Behavior

Retryable jobs will always retry irrespective of unique configuration

[sorentwo]

As reported, this issue only impacts Pro v1.5 and will be "bandaged" in v1.5.3. The solution we'd like to drive toward is providing a set of ordained unique states that are safe to use and not inherently flawed. That can either be through functions on Oban.Job, or through named expansions.

The named expansions would be used like unique: [states: :incomplete], where the expansions are:

• :all - ~w(available scheduled executing retryable complete cancelled discarded)a
• :incomplete - ~w(available scheduled executing retryable)a
• :scheduled - ~w(scheduled)
• :successful - ~w(available scheduled executing retryable complete)a (the default)

Setting a list of states would be deprecated, but continue to work for backward compatibility. The obvious outlier above is for scheduled, which is safe by itself and useful for debounced insertions.