Brief approach: the leak is caused by a global container accumulating session objects with no eviction. Fix by changing the container to an eviction-capable structure (LRU/TTL) or weak references and adding explicit lifecycle management. Detect with tracemalloc/objgraph and mitigate in prod via memory limits, rolling restarts, and live profiling.Code fixes (examples):Use collections.OrderedDict TTL/LRU:python
from collections import OrderedDict
import time
class TTLCache:
def __init__(self, max_size=10000, ttl=3600):
self.store = OrderedDict()
self.max_size = max_size
self.ttl = ttl
def set(self, key, value):
now = time.time()
if key in self.store:
self.store.pop(key)
self.store[key] = (value, now)
if len(self.store) > self.max_size:
self.store.popitem(last=False) # evict oldest
def get(self, key):
val = self.store.get(key)
if not val:
return None
value, created = val
if time.time() - created > self.ttl:
self.store.pop(key, None)
return None
return value
python
import weakref
_session_map = weakref.WeakValueDictionary() # entries auto-removed when object dies
