A table is a promise about the data underneath it: more than one dimension here, worth a row and a column each. Every domain picker built earlier the same day kept that promise honestly, with one column for a name and one for a link. It was correct, and also heavier than the one thing it was saying: pick one of these.
What a table costs when it only has one real column
A table announces rows and columns before you’ve read either. It reserves a card, a header bar and a border, then delivers one fact per row: a name. The “Manage” link in the second column wasn’t information; it was the same action repeated per row. Twelve pages had independently arrived at the same shape, which usually means the shape itself is asking for revision.
Consolidating without assuming uniformity
The instinct on noticing twelve nearly identical blocks is to collapse them all. That is right for eleven and wrong for the twelfth, which showed a status badge, an issuer and an expiry date next to each domain, real information a plain dropdown has no room for. Folding it in would have deleted data for the sake of consistency. Noticing the exception mattered more than executing the rule.
The one page that needed a different shape entirely
A second page needed a different control, not a lighter one. Elsewhere “pick a domain” narrows the view to one thing. On this page the most useful view is everything from every domain together, with a domain-scoped view as the exception. A dropdown can’t express “and also show me the combined view”, but tabs can, with “all of it” as a first-class choice.
What stayed the same on purpose
Nothing about how a choice is submitted changed: which domain a new record belongs to, and the check that you may touch the one you picked, are as before. Only the part a person looks at changed.