When to Denormalize

Denormalization is a legitimate technique and a terrible default.

It is legitimate when you have measured a real performance problem, the join is genuinely the cause, and you have a plan for keeping the copy correct: a trigger, a scheduled job, or accepting staleness deliberately and saying how stale.

It is a problem when it happens by accident, which is what the flat table above is.

The honest ordering. Normalize first. Measure. Denormalize the specific thing the measurement pointed at, and write down how the copy stays correct.

A denormalized column with no answer to 'what updates this' is a column that will be wrong within the year.