Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@README.md`:
- Line 184: Update the test count in the README status line to match the
canonical collected-test total: run the canonical source (python
scripts/check_doc_counts.py or pytest tests/ --collect-only -q) to regenerate
the accurate test number, then replace the hardcoded "3,729 tests" string in the
README.md status sentence with the new value so the public doc matches the
authoritative script output.

In `@vera/wasm/closures.py`:
- Around line 142-156: The current logic in the call-site that builds
arg_wasm_types uses _infer_expr_wasm_type and treats None as "Unit" implicitly;
make this defensive by asserting or checking that any None result indeed
corresponds to a Unit-typed parameter from the closure signature before skipping
it (or raise a clear error if not). Update the code around the arg handling in
the same function that populates arg_wasm_types (the apply_fn translation path
where wt = self._infer_expr_wasm_type(arg) and the i32_pair handling occurs) to
verify the parameter type (or add an explicit assert/comment referencing that
closure-lift already skips Unit params) so unexpected None values are caught
with a clear message rather than silently ignored.

---

Outside diff comments:
In `@TESTING.md`:
- Line 86: Update the stale counts in TESTING.md to match the canonical numbers
from scripts/check_doc_counts.py: run scripts/check_doc_counts.py to obtain the
current totals, then replace the “84” in the conformance suite sentence (the
phrase describing the number of programs) and the “410 tests” occurrence (lines
referencing the total test count) so they match the script output (e.g., change
to “85” and the updated total such as “425” if that is what the script reports);
ensure both the single-program count (conformance suite description) and the
total tests sentence are consistent with scripts/check_doc_counts.py.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@TESTING.md`:
- Around line 9-11: Update the two stale counts in TESTING.md to match the
already-updated values: change the Conformance Suite intro count that currently
reads "84" to "85" (match the table entry "Conformance programs | 85 programs
across 9 spec chapters") and update the pytest example that currently says "410
tests" to "425 tests" so the document is self-consistent with the table and the
expected pytest/ scripts output.

---

Duplicate comments:
In `@README.md`:
- Line 184: The README's public status line currently hardcodes "3,729 tests"
and is out of sync; run the authoritative count script (python
scripts/check_doc_counts.py) or collect tests via pytest (pytest tests/
--collect-only -q) to regenerate the tests total, then update the README
occurrence of "3,729 tests" to the new number returned by those tools so the
documented test count matches the real collected total.

In `@vera/wasm/closures.py`:
- Around line 144-156: The code incorrectly treats any None result from
_infer_expr_wasm_type as Unit and skips it when building arg_wasm_types; change
the logic to explicitly check the closure parameter's declared type (the
parameter node/type used when building the signature) and only skip when that
declared type is Unit, or alternatively call a helper that returns Unit-only
semantics, ensuring you still expand "i32_pair" into ["i32","i32"] and append
other concrete wasm types; update the logic around _infer_expr_wasm_type,
arg_wasm_types, and the closure signature construction so that only true Unit
parameters are omitted from the call_indirect signature.