[v1.5] Manifest overwritten in subsequent parses, which can affect "stateful" worfklows

[jtcohen6]

See closed dbt-core issue linked below for context. We should also add a note to v1.5 migration guide.

---

          @jtcohen6 Thanks for the reply. I thought this was a pretty run-of-the-mill workflow, (Steps 3 and 4, anyway).

Is there a more typical development / Slim CI pattern for selecting modified nodes that does not require this temp folder workaround? More to the point - is there a practical advantage to updating the model checksum during parse, as opposed to later in the pipeline?

Anyway, I appreciate the feedback; this is enough to fix my CI, and I suppose I can just make a script for development. However, I think this change should be captured in one of these doc pages/sections:

• https://docs.getdbt.com/reference/node-selection/state-comparison-caveats
• https://docs.getdbt.com/reference/node-selection/methods#the-state-method

Originally posted by @serene-capybara in https://github.com/dbt-labs/dbt-core/issues/7790#issuecomment-1578806524

[mirnawong1]

Hi @jtcohen6 and @matthewshaver - given the 1.5 tag, is this still open?