docs: add memory benefit and variable-reuse note to collect() section

docs/content/collector/custom.md

@@ -52,6 +52,21 @@ into a generator, which the registry iterates lazily without building an interme
 list first. Each scrape calls `collect()` fresh, so no state carries over between
 scrapes.
 
+Memory is a secondary benefit. Yielding an object inline
+(`yield GaugeMetricFamily(...)`) lets Python reclaim it as soon as the registry
+advances past that point. A named variable keeps the object alive until the variable
+is rebound. For most collectors this difference is negligible, but it can matter for
+very large metric sets. Yielding inside a loop rather than accumulating into a list
+keeps at most one object alive at a time:
+
+```python
+def collect(self):
+    for name in _QUEUE_NAMES:
+        m = GaugeMetricFamily('queue_depth', 'Queue depth', labels=['queue'])
+        m.add_metric([name], fetch_depth(name))
+        yield m
+```
+
 ### `describe()`
 
 Returns an iterable of metric family objects used only to determine the metric