Summarize Support-Ticket Root Causes

#CONTEXT:
Summarize recurring support-ticket themes and possible root causes for human review. Analyze only the supplied corpus and rules. This is an evidence review, not a final causal finding or authorization to change operations.

#INPUTS:
- Authorized, minimized ticket corpus with stable ticket references: [SUPPORT TICKET CORPUS]
- Start date, end date, timezone, and inclusion rules: [ANALYSIS WINDOW]
- Approved issue categories and definitions: [CATEGORY DEFINITIONS]
- Approved severity levels and criteria: [SEVERITY RULES]
- Supplied team, queue, product, and process ownership data: [OWNERSHIP DATA]
- Verified releases, process changes, outages, or other known events: [KNOWN CHANGES]
- Minimum ticket count and evidence quality required for a root-cause hypothesis: [MINIMUM EVIDENCE THRESHOLD]

#INSTRUCTIONS:
1. Validate the corpus boundary. Report duplicates, missing dates, conflicting fields, sampling limits, and Unknown coverage before analysis.
2. Do not invent, infer, or fabricate policy, eligibility, customer facts, outcomes, ticket content, counts, owner responsibility, fault, or root cause.
3. Classify tickets only against the supplied category and severity definitions. Keep unmatched or ambiguous tickets in an Unknown or conflict group.
4. Produce exact counts and percentages using the included corpus as the denominator. Show the calculation and never generalize beyond the analysis window.
5. Separate observed symptoms, recurring themes, correlations, and possible root-cause hypotheses. A hypothesis must meet the supplied evidence threshold and cite ticket references plus any known change; otherwise label it Unknown or insufficient evidence.
6. Do not infer customer intent, employee performance, legal responsibility, or fault from ticket language or correlations.
7. Recommend investigation questions and internal evidence to collect. Do not route, assign, close, resolve, delete, archive, or update a ticket, account, record, or system.
8. Do not issue or recommend an automatic refund, credit, compensation, retention offer, discount, exception, admission, or customer communication based on aggregate themes.
9. Use only authorized ticket data and authorized customer data in an approved AI workspace or approved AI vendor. Minimize the data and redact or omit unnecessary personal, sensitive, or confidential data. Follow company retention and vendor policy.
10. A named human owner must approve every refund, credit, or compensation decision before action. A named human owner and legal counsel must approve every admission of fault or liability before action. A named human owner must approve every retention offer, discount, or policy exception before action. A named human owner must approve every external communication, sending action, or customer contact before action. A named human owner must approve every ticket, system, account, or record change or mutation before action.

#RESPONSE FORMAT:
## Corpus quality and Unknowns
- Included and excluded records
- Duplicate, missing, and conflicting data
- Analysis limitations

## Evidence-based root-cause themes
| Theme | Ticket count | Percentage | Evidence references | Symptom or hypothesis | Threshold status | Confidence limit |
|---|---:|---:|---|---|---|---|

## Unresolved hypotheses
- Hypothesis — missing evidence — investigation question

## Human-reviewed investigation plan
| Proposed investigation | Named owner | Evidence needed | Prohibited action |
|---|---|---|---|