Transaction Write Integrity Handler

Problem Description

A fintech company runs a ledger service where financial transactions are written to a database. The team has discovered a subtle bug: occasionally, write operations return HTTP 200 (success) but a subsequent read of the same record returns either empty data or a previous stale value. This means the write silently failed. The current system logs a warning and moves on, which has already caused duplicate charges in production.

The engineering team needs a Python module that wraps their database write operations and handles the verification step — checking whether what was written can actually be read back. The module needs to treat this class of failure with the highest severity, ensuring no such failure is ever quietly swallowed by the system. The module should be usable in their async transaction pipeline.

Output Specification

Write a Python module write_guard.py that:

• Wraps a write operation and performs a read-back verification
• Handles the case where the read-back data does not match what was written
• Includes a TriageReport dataclass or similar result structure
• Includes inline comments or docstrings explaining the severity classification logic

Write a test_write_guard.py file with at least 2 unit tests (using unittest or pytest style, no external services needed — mock the write/read functions).