Self-review in the fee-logic generator session keeps defending its own choices. What limitation is showing?

When self-review inside the fee-logic generator session keeps defending its own choices, the limitation on display is that retained generation reasoning context biases the critique. The same context that produced the patch supplies justifications the reviewer-model reuses instead of independently stress-testing edge cases in municipal fee ordinances. Self-review is therefore weaker for catching subtle logic bugs than an independent instance. Claiming self-review is always stronger than any independent instance reverses the usual bias finding. Asserting retained generation context never affects critique quality ignores why defense patterns appear. Treating defense of prior choices as proof the patch needs no further review confuses rhetorical confidence with correctness and is especially risky for resident-facing charges. Exam caveat: self-review can still catch surface issues; do not rely on it as the sole gate for fee, tax, or eligibility logic. Operational check: when a generating session's self-critique mostly restates its plan, spawn an independent reviewer without that chain-of-thought, compare finding sets, and prefer independent findings for merge decisions on fee-logic PRs.