diff --git a/include/proto/stick_table.h b/include/proto/stick_table.h
index d9a25d0..0c26fbe 100644
--- a/include/proto/stick_table.h
+++ b/include/proto/stick_table.h
@@ -52,6 +52,7 @@
 int stktable_compatible_sample(struct sample_expr *expr, unsigned long table_type);
 int stktable_get_data_type(char *name);
 struct proxy *find_stktable(const char *name);
+int stktable_trash_oldest(struct stktable *t, int to_batch);
 
 /* return allocation size for standard data type <type> */
 static inline int stktable_type_size(int type)
diff --git a/include/types/stick_table.h b/include/types/stick_table.h
index dcdc405..e28492c 100644
--- a/include/types/stick_table.h
+++ b/include/types/stick_table.h
@@ -150,6 +150,7 @@
 	struct task *sync_task;   /* sync task */
 	unsigned int update;
 	unsigned int localupdate;
+	unsigned int syncing;     /* number of sync tasks watching this table now */
 	union {
 		struct peers *p; /* sync peers */
 		char *name;
diff --git a/src/peers.c b/src/peers.c
index 83781ba..7247d76 100644
--- a/src/peers.c
+++ b/src/peers.c
@@ -1401,6 +1401,7 @@
 				/* add DO NOT STOP flag if not present */
 				jobs++;
 				st->flags |= SHTABLE_F_DONOTSTOP;
+				st->table->syncing++;
 			}
 
 			/* disconnect all connected peers */
@@ -1418,6 +1419,7 @@
 				/* resync of new process was complete, current process can die now */
 				jobs--;
 				st->flags &= ~SHTABLE_F_DONOTSTOP;
+				st->table->syncing--;
 			}
 		}
 		else if (!ps->session) {
@@ -1440,6 +1442,7 @@
 					/* unable to resync new process, current process can die now */
 					jobs--;
 					st->flags &= ~SHTABLE_F_DONOTSTOP;
+					st->table->syncing--;
 				}
 			}
 		}
diff --git a/src/proxy.c b/src/proxy.c
index b67f024..37bda48 100644
--- a/src/proxy.c
+++ b/src/proxy.c
@@ -560,6 +560,24 @@
 		}
 	}
 
+	/* If the proxy holds a stick table, we need to purge all unused
+	 * entries. These are all the ones in the table with ref_cnt == 0
+	 * and all the ones in the pool used to allocate new entries. Any
+	 * entry attached to an existing session waiting for a store will
+	 * be in neither list. Any entry being dumped will have ref_cnt > 0.
+	 * However we protect tables that are being synced to peers.
+	 */
+	if (unlikely(stopping && p->state == PR_STSTOPPED && p->table.current)) {
+		if (!p->table.syncing) {
+			stktable_trash_oldest(&p->table, p->table.current);
+			pool_gc2();
+		}
+		if (p->table.current) {
+			/* some entries still remain, let's recheck in one second */
+			next = tick_first(next, tick_add(now_ms, 1000));
+		}
+	}
+
 	/* the rest below is just for frontends */
 	if (!(p->cap & PR_CAP_FE))
 		goto out;
diff --git a/src/stick_table.c b/src/stick_table.c
index c357b32..b1d8827 100644
--- a/src/stick_table.c
+++ b/src/stick_table.c
@@ -90,7 +90,7 @@
  * Trash oldest <to_batch> sticky sessions from table <t>
  * Returns number of trashed sticky sessions.
  */
-static int stktable_trash_oldest(struct stktable *t, int to_batch)
+int stktable_trash_oldest(struct stktable *t, int to_batch)
 {
 	struct stksess *ts;
 	struct eb32_node *eb;