Troubleshooting a Zap that runs but returns incorrect results

We’re working with some genuinely brilliant people in Zapier Support on a workflow issue at the moment, and before we put yet another round of questions back to them, I wanted to ask the community whether anyone has seen something similar and might have a useful technical suggestion.

We are trying to avoid wasting Support’s time if this turns out to be something obvious that an experienced builder would spot immediately.

The situation, in simple terms, is this:

We have a member-matching workflow where the Zap can appear to complete mechanically, but the actual result is still wrong in practical terms. In other words:

• steps can run

• an email can send

• a record can update

• but the output can still come back with zero / fallback-style results when data should exist

The main technical themes we are trying to understand are:

1. Draft vs published mismatch
   There are signs that the draft version and the actually executing published version may not have been aligned, which may explain why changes that looked correct in the editor did not produce the expected live behaviour.

2. Field-reference mismatch
   A key field appears to behave differently depending on whether the raw field or the __value version is used, especially in path filters.

3. Code-step input / payload-shape inconsistency
   At one stage the Code-step inputs were invalid and had to be re-mapped using Step Output fields. That helped, but the returned result was still zero, which suggests that fixing the mapping alone was not the whole answer.

4. Zapier Tables UI vs backend behaviour
   Separately, we have also seen a case where records appeared present in backend counts and in retrieval, but were not rendering properly across later pages of the Tables UI.

We are not looking to criticise Zapier Support at all. Quite the opposite. They have been very engaged and are actively helping us. I am only asking here in case someone in the community has seen this exact pattern before and can save everyone some time.

The specific questions are:

• Have you seen cases where the published workflow was effectively behind the draft and caused troubleshooting to become misleading?

• Have you seen path filters behave differently depending on raw field vs __value usage?

• If a Code step is receiving Step Output data but still returning zero / fallback-style output, what would you inspect first?

• Have you seen Zapier Tables show records as effectively present in retrieval, but not render them properly in the UI?

Any practical ideas would be very welcome. We are simply trying to sense-check whether there is anything obvious we may have missed before putting more time back onto the Support team.