[dgafka]

Why is this change proposed?

Why

Flow conformance stages that consume more than one message now keep polling until their queue is empty, so a slow CI runner no longer fails them.

These stages polled with ExecutionPollingMetadata::createWithTestingSetup(), which stops the consumer after any poll that ends past 100ms. When the first message took longer than that, the consumer stopped before the second message and the expected span tree was missing. This failed Split Testing (8.5) on #727 (job): "dynamic channel sends round robin to its channels" recorded the send to orders_b but no receive under it. The push run on the same commit passed.

Before / after

Stage needs	Before	After
A second poll, and the first message took over 100ms	Consumer stops early; the conformance assertion fails intermittently	Consumer polls until the queue is empty
One message only	Unchanged	Unchanged

Description of Changes

• Five catalogue stages now poll with createWithFinishWhenNoMessages(): round robin across two channels, two commands on one queue, a handler feeding its own channel, retry then dead letter, and a failing error-channel send.
• Only test scenarios changed; production code is unchanged.
• The cause was confirmed by forcing the limit to 1ms across the catalogue: exactly these five stages failed, and every other stage handles one message per poll and still passes.

Pull Request Contribution Terms

• I have read and agree to the contribution terms outlined in CONTRIBUTING.