Good questions — and what you’re seeing is consistent with how this feature is currently scoped.
- Tenant-based column aliases are scoped at the Org/tenant level, not by Group.
I could not find official documentation support for group-level scoping of tenant column aliases via CSV or TML.
- If your update included fields that are not valid/eligible in that context, it can result in a partial set appearing on download (only recognized/accepted alias entries show up).
On the doc statement:
> “For published Models, the alias must be defined in the primary Org.”
Based on the documented behavior, this means:
- If a Model is published (shared as published content), alias resolution expects the alias definition to exist in the primary Org.
- So yes, for that published-model path, primary Org definition is required.
- If your models only exist outside primary Org, that likely explains why your alias behavior is inconsistent with expectations for published content.
Also, a TML-related gotcha: in ThoughtSpot modeling, column identifiers in connected objects depend on the `name`/identifier chain, so edits can behave unexpectedly if upstream/downstream identifiers don’t align exactly across related TML objects.
If you want, I can help you validate a concrete migration path (primary-Org alias definition + tenant/org distribution model) and a checklist to confirm why specific fields are dropped from your translation export.