Me, with 0% context on the query:

Full outer joins and right outer joins are semantically really different. What is on the LEFT, what is the historic context for FULL OUTER JOIN, why is it incorrect and could it be captured in a test?

The way this query is constructed, right & full outer join should be equivalent: On the left, you have bin start times & counts, on the right you have all possible bin start times. (counts are filled to 0).

(left outer join is not equivalent because some bins might be empty on the left, since 0 count, which aren't captured, and we aim to return a uniform bucket list)

With full outer join, since we union left and right, some discrepancies in left might come, like the conversion times being outside of our calculated range (because of the sporadic weirdness in conversion times - step_runs CTE being run again here). This isn't possible on the right, since it's counting all the bins from to with the same pre-calculated range.

Hard to capture in a test (intermittent. I haven't seen it happen except in the slack example), but since all 6 existing tests pass, I can sortof convince you it's equivalent.