objectql: the bare-name refusal still explains itself with the old one-name-per-deployment rule for the four types that now hold one name per namespace #33507

domain:server#7 · os-tesla · session_01L9qibzuZJBQY4wUQFJbkfx

Filing gate: found by the #31699 card thread while it fixed PR #33431's dogfood pins. An administrator whose install is refused reads a reason that is no longer true, and so cannot tell what to rename. ⛔ Not a claim; triage sets the lane and grade.

What happens today

bareNameReason in packages/objectql/src/bare-name-one-holder.ts (around line 621) builds the sentence every bare-name refusal and boot warning carries. For every type it says the type "holds one name per deployment … with no package to tell two definitions apart, so with two holders every caller is served the same one, chosen by registration order."

That is still true for the types that have not narrowed. Flow (#31696), webhook (#32577), email template (#32214) and, once #33431 lands, connector now hold one name per namespace, and two holders are refused naming both rather than served by registration order. A co-owner refused for one of these four types (for example, a second package of alpha's namespace declaring alpha's connector name) is given the old reason. This predates #33431; main has said it for the flow since #31696.

What this card would build

• Give the four namespace-scoped types (NAMESPACE_SCOPED_BARE_NAME_TYPES) a reason that states their rule: one holder per namespace; another namespace may ship the name; a bare name two holders share is refused naming both.
• Keep today's sentence for the types still held one per deployment.
• Pin the sentence for one narrowed type and one unnarrowed type. Move any dogfood or unit pin that quotes the old text.

Seam

Seam: objectql:bare-name-one-holder.ts (bareNameReason, NAMESPACE_SCOPED_BARE_NAME_TYPES).

Dedupe words: bareNameReason, one name per deployment, chosen by registration order, bare-name refusal reason namespace.