Choosing a boundary, a timeout, a queue, or a storage model is a prediction. It says something about traffic, latency, failure, ownership, or the people who will operate the result.
Some predictions are cheap to change. A configuration value can move. A boundary that has accumulated data, dependencies, and expectations is harder to move.
Confidence is not the absence of uncertainty. It is a record of what the uncertainty is resting on.
The useful question is not whether an architecture is certain. It is what we are assuming, what evidence supports it, and how difficult the decision would be to reverse if the prediction is wrong.
An architecture decision earns confidence when its consequences are visible before the system makes them expensive.