Address review round 2: self-heal corrupt cache entries; comment accuracy

- get_cached_result: explicitly catch decode/JSON errors from a corrupt or
  truncated gz: entry, purge it, and return None so it repopulates on the next
  call (was swallowed by the outer handler -> permanent miss). (Copilot)
- _decode_cache_field: reword the comment so it no longer implies the function
  itself purges (the caller does), and fix "un-caught" -> "uncaught". (Copilot)
- test: a corrupt gz: payload decodes to the raw string without raising.

Author: Robbie1977

src/vfbquery/solr_result_cache.py

@@ -68,9 +68,10 @@ def _decode_cache_field(cached_field) -> str:
             blob = base64.b64decode(cached_field[len(_CACHE_GZIP_PREFIX):])
             return gzip.decompress(blob).decode("utf-8")
         except Exception:
-            # Corrupt/truncated gz payload: return the raw string so callers'
-            # json.loads fails and the entry is treated as invalid (and purged),
-            # rather than raising an un-caught error that aborts cleanup/stats.
+            # Corrupt/truncated gz payload: return the raw string rather than
+            # raising an uncaught error that would abort cleanup/stats runs. The
+            # caller's json.loads then fails, so the entry is treated as invalid
+            # JSON (get_cached_result purges it; other callers skip it).
             logger.warning("Failed to decode compressed cache payload; treating as invalid")
             return cached_field
     return cached_field
@@ -309,8 +310,15 @@ def get_cached_result(self, query_type: str, term_id: str, **params) -> Optional
             if isinstance(cached_field, list):
                 cached_field = cached_field[0]
 
-            # Parse the cached metadata and result
-            cached_data = json.loads(_decode_cache_field(cached_field))
+            # Parse the cached metadata and result. A corrupt/undecodable entry
+            # (e.g. a truncated gz: payload) raises here; purge it so the next
+            # call repopulates, rather than leaving a permanent cache miss.
+            try:
+                cached_data = json.loads(_decode_cache_field(cached_field))
+            except (ValueError, TypeError):
+                logger.warning(f"Corrupt cache entry for {query_type}({term_id}); clearing it")
+                self._clear_expired_cache_document(cache_doc_id)
+                return None
 
             # Check package version before anything else so stale cache is rejected early.
             # Only invalidate when the cached entry is OLDER than the current code

tests/test_gzip_cache.py

@@ -13,6 +13,13 @@ def test_roundtrip_compresses_and_restores():
     assert _decode_cache_field(enc) == env
 
 
+def test_decode_returns_raw_string_on_corrupt_gz_payload():
+    # A gz:-prefixed but undecodable value must not raise; it returns the raw
+    # string so the caller's json.loads fails and the entry is treated as invalid.
+    for bad in (_CACHE_GZIP_PREFIX + "!!!not-base64!!!", _CACHE_GZIP_PREFIX + "AAAA"):
+        assert _decode_cache_field(bad) == bad
+
+
 def test_decode_handles_legacy_plain_json_and_list_shape():
     legacy = json.dumps({"result": 1})
     assert _decode_cache_field(legacy) == legacy