Turn a spike in similar tickets into a root cause writeup someone can actually act on

You are a support operations analyst investigating why a specific type of ticket has been recurring. I am not asking you to guess a cause from nothing — I will give you the pattern as it looks from the outside and a sample of real ticket language, and your job is to separate what customers are reporting (the symptom) from what is actually causing it (the root cause), which are very often not the same thing.

TICKET PATTERN
Since last Tuesday, roughly 30 tickets a day from customers on the annual plan saying their exported reports are missing the last week of data.

SAMPLE TICKET EXCERPTS
"My Monday export only goes up to the 14th, but the dashboard shows data through the 21st." / "Support told me last time this was a caching issue but it's back again."

CURRENT HYPOTHESIS (if any)
Engineering thinks it's a timezone bug in the nightly export job.

BUSINESS CONTEXT
We shipped a new data-warehouse migration nine days ago; the export feature wasn't explicitly called out in that release's changelog.

STEP 1 — SYMPTOM VS CAUSE
Read the sample excerpts and state, in one sentence, what customers believe is wrong. Then list at least three distinct places upstream where that symptom could actually originate (a product bug, a policy that was never explained, a third-party dependency, a documentation gap, an onboarding step that was skipped) — do not stop at the first plausible cause, because a root cause analysis that settles on the most obvious answer is usually just re-describing the symptom with more technical vocabulary.

STEP 2 — TEST THE CURRENT HYPOTHESIS
If a current hypothesis was given, treat it as one candidate among several, not the answer — state what evidence in the excerpts supports it, what evidence is missing that would be needed to confirm it, and what a wrong-but-plausible-looking hypothesis would look like here so I don't anchor on the first guess that was already in the room before this analysis started.

STEP 3 — RANK THE CANDIDATES
Rank the candidate causes by how well they explain every excerpt I gave you, not just the most dramatic one — a real root cause should explain the mundane tickets in the batch as well as the angry one, and a candidate that only explains one excerpt is probably a contributing factor, not the root cause.

STEP 4 — WHAT WOULD CONFIRM IT
For the top-ranked candidate, state exactly what evidence (a log query, a specific customer to re-contact, a specific screen to check) would confirm or rule it out — never present a ranked guess as if it were already confirmed.

WHAT NOT TO DO
Do not recommend a fix before the cause is confirmed. Do not invent specific error codes, system names, or statistics that weren't in what I gave you — if you need a detail to test a hypothesis, say so as an open question rather than filling it in.

OUTPUT FORMAT
1. One-sentence symptom statement.
2. Ranked table of candidate root causes with the evidence for/against each.
3. The single most likely root cause and why it beats the others.
4. The specific next step that would confirm it.