[zzylol]

Closes #112.

The coverage floors in promql_corpus.rs were set from figures that predate the #40 / #43–#51 / #84 / #86 / #89 coverage work.

	comment claimed	actual on main
docs lowered	≈29	48 (of 49)
testdata lowered	≈574	1502 (of 1823)
testdata rejected	1014	86
testdata unparseable	235	235 ✓

td.lowered >= 520 therefore passed with ~980 queries of slack: lowering could have broken for 65% of the testdata corpus without turning the test red.

Change

Floors raised to ~1% under the measured values (docs >= 47, td >= 1485), and the stale comment replaced with the current numbers plus a pointer to what the residual failures are (#108 parser gaps, #109 lowering gaps).

Verification

Confirmed the assert is actually live, rather than assuming it:

# floor temporarily set to 1503 (one above actual)
testdata lowering coverage regressed: Tally { lowered: 1502, rejected: 86, unparseable: 235 }
test result: FAILED

# restored to 1485
test result: ok

cargo clippy --all-targets clean.

🤖 Generated with Claude Code