sidekiq-unique-jobs: Jobs executing and immediately returning

Describe the bug We’ve had two instances in which we were using sidekiq_options unique: :until_and_while_executing on a worker and after some time, calling .perform_async on the worker would result in the job getting enqueued, a job ID assigned, picked up by a worker, and then exit immediately.

For example, we would see in the sidekiq.log

2019-09-09T14:52:16.545Z 14236 TID-ovo800gss MyWorker JID-3287686b0cad75fee25a57c8 INFO: start
2019-09-09T14:52:16.548Z 14236 TID-ovo800gss MyWorker JID-3287686b0cad75fee25a57c8 INFO: done: 0.003 sec

In all cases, there were no other instances of MyWorker enqueued or running.

Upon changing to sidekiq_options unique: :until_executing the problem went away in both cases.

Expected behavior The worker to execute as normal.

Current behavior See above. Note in the class below we are logging start, and success or failure - in these cases we see nothing logged. These are also typically workers that run for several minutes.

Worker class

class MyWorker
  include Sidekiq::Worker

  sidekiq_options queue: :data_sync, retry: true, backtrace: true
  sidekiq_options unique: :until_and_while_executing

  def perform(some_id)
    some_model = SomeModel.find(some_id)

    log('Started', some_model)
    result = some_model.sync.call

    if result.success?
      log('Finished', some_model)
    else
      log("Failed: #{result.errors.join(', ')}", some_model)
    end
  end

  private

  def log(message, some_model = nil)
    EventLogger.log(
      :sync,
      note: message,
      loggable: some_model
    )
  end
end

Additional context Anything we can look for here? Ideally we’d want to use :until_and_while_executing if possible