diff --git a/src/basic/unit-def.c b/src/basic/unit-def.c
index 86b66e2be0..bfd748ada4 100644
--- a/src/basic/unit-def.c
+++ b/src/basic/unit-def.c
@@ -152,6 +152,7 @@ static const char* const mount_state_table[_MOUNT_STATE_MAX] = {
         [MOUNT_REMOUNTING_SIGKILL] = "remounting-sigkill",
         [MOUNT_UNMOUNTING_SIGTERM] = "unmounting-sigterm",
         [MOUNT_UNMOUNTING_SIGKILL] = "unmounting-sigkill",
+        [MOUNT_UNMOUNTING_CATCHUP] = "unmounting-catchup",
         [MOUNT_FAILED]             = "failed",
         [MOUNT_CLEANING]           = "cleaning",
 };
diff --git a/src/basic/unit-def.h b/src/basic/unit-def.h
index 169e1f719e..7bd86cb1d2 100644
--- a/src/basic/unit-def.h
+++ b/src/basic/unit-def.h
@@ -97,6 +97,7 @@ typedef enum MountState {
         MOUNT_REMOUNTING_SIGKILL,
         MOUNT_UNMOUNTING_SIGTERM,
         MOUNT_UNMOUNTING_SIGKILL,
+        MOUNT_UNMOUNTING_CATCHUP,
         MOUNT_FAILED,
         MOUNT_CLEANING,
         _MOUNT_STATE_MAX,
diff --git a/src/core/mount.c b/src/core/mount.c
index 4a0dc7ebac..295e4c66c7 100644
--- a/src/core/mount.c
+++ b/src/core/mount.c
@@ -48,6 +48,7 @@ static const UnitActiveState state_translation_table[_MOUNT_STATE_MAX] = {
         [MOUNT_REMOUNTING_SIGKILL] = UNIT_RELOADING,
         [MOUNT_UNMOUNTING_SIGTERM] = UNIT_DEACTIVATING,
         [MOUNT_UNMOUNTING_SIGKILL] = UNIT_DEACTIVATING,
+        [MOUNT_UNMOUNTING_CATCHUP] = UNIT_DEACTIVATING,
         [MOUNT_FAILED] = UNIT_FAILED,
         [MOUNT_CLEANING] = UNIT_MAINTENANCE,
 };
@@ -1070,11 +1071,37 @@ static void mount_enter_unmounting(Mount *m) {
 
         assert(m);
 
+        r = path_is_mount_point(m->where, NULL, 0);
+        if (IN_SET(r, 0, -ENOENT)) {
+                /* Not a mount point anymore ? Then we raced against something else which unmounted it? Let's
+                 * handle this gracefully, and wait until we get the kernel notification about it. */
+
+                log_unit_debug(UNIT(m), "Path '%s' is not a mount point anymore, waiting for mount table refresh.", m->where);
+
+                /* Apparently our idea of the kernel mount table is out of date. Make sure we re-read it
+                 * again, soon, so that we don't delay mount handling unnecessarily. */
+                (void) sd_event_source_leave_ratelimit(UNIT(m)->manager->mount_event_source);
+
+                m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID;
+                mount_unwatch_control_pid(m);
+
+                r = mount_arm_timer(m, usec_add(now(CLOCK_MONOTONIC), m->timeout_usec));
+                if (r < 0)
+                        goto fail;
+
+                /* Let's enter a distinct state where we just wait for the mount table to catch up */
+                mount_set_state(m, MOUNT_UNMOUNTING_CATCHUP);
+                return;
+        }
+        if (r < 0)
+                log_unit_debug_errno(UNIT(m), r, "Unable to determine if '%s' is a mount point, ignoring: %m", m->where);
+
         /* Start counting our attempts */
         if (!IN_SET(m->state,
                     MOUNT_UNMOUNTING,
                     MOUNT_UNMOUNTING_SIGTERM,
-                    MOUNT_UNMOUNTING_SIGKILL))
+                    MOUNT_UNMOUNTING_SIGKILL,
+                    MOUNT_UNMOUNTING_CATCHUP))
                 m->n_retry_umount = 0;
 
         m->control_command_id = MOUNT_EXEC_UNMOUNT;
@@ -1095,7 +1122,6 @@ static void mount_enter_unmounting(Mount *m) {
                 goto fail;
 
         mount_set_state(m, MOUNT_UNMOUNTING);
-
         return;
 
 fail:
@@ -1268,6 +1294,7 @@ static int mount_start(Unit *u) {
                    MOUNT_UNMOUNTING,
                    MOUNT_UNMOUNTING_SIGTERM,
                    MOUNT_UNMOUNTING_SIGKILL,
+                   MOUNT_UNMOUNTING_CATCHUP,
                    MOUNT_CLEANING))
                 return -EAGAIN;
 
@@ -1297,6 +1324,7 @@ static int mount_stop(Unit *u) {
         case MOUNT_UNMOUNTING:
         case MOUNT_UNMOUNTING_SIGKILL:
         case MOUNT_UNMOUNTING_SIGTERM:
+        case MOUNT_UNMOUNTING_CATCHUP:
                 /* Already on it */
                 return 0;
 
@@ -1562,6 +1590,10 @@ static void mount_sigchld_event(Unit *u, pid_t pid, int code, int status) {
                 mount_enter_dead_or_mounted(m, f);
                 break;
 
+        case MOUNT_UNMOUNTING_CATCHUP:
+                log_unit_debug(u, "Was notified about control process death, but wasn't expecting it. Ignoring.");
+                break;
+
         case MOUNT_CLEANING:
                 if (m->clean_result == MOUNT_SUCCESS)
                         m->clean_result = f;
@@ -1636,6 +1668,11 @@ static int mount_dispatch_timer(sd_event_source *source, usec_t usec, void *user
                 mount_enter_dead_or_mounted(m, MOUNT_FAILURE_TIMEOUT);
                 break;
 
+        case MOUNT_UNMOUNTING_CATCHUP:
+                log_unit_warning(UNIT(m), "Waiting for unmount notification timed out. Giving up.");
+                mount_enter_dead_or_mounted(m, MOUNT_FAILURE_TIMEOUT);
+                break;
+
         case MOUNT_CLEANING:
                 log_unit_warning(UNIT(m), "Cleaning timed out. killing.");
 
@@ -2077,8 +2114,7 @@ static int mount_process_proc_self_mountinfo(Manager *m) {
 
                 if (!mount_is_mounted(mount)) {
 
-                        /* A mount point is not around right now. It
-                         * might be gone, or might never have
+                        /* A mount point is not around right now. It might be gone, or might never have
                          * existed. */
 
                         if (mount->from_proc_self_mountinfo &&
@@ -2093,6 +2129,7 @@ static int mount_process_proc_self_mountinfo(Manager *m) {
                         switch (mount->state) {
 
                         case MOUNT_MOUNTED:
+                        case MOUNT_UNMOUNTING_CATCHUP:
                                 /* This has just been unmounted by somebody else, follow the state change. */
                                 mount_enter_dead(mount, MOUNT_SUCCESS);
                                 break;
@@ -2131,11 +2168,9 @@ static int mount_process_proc_self_mountinfo(Manager *m) {
                                 break;
 
                         default:
-                                /* Nothing really changed, but let's
-                                 * issue an notification call
-                                 * nonetheless, in case somebody is
-                                 * waiting for this. (e.g. file system
-                                 * ro/rw remounts.) */
+                                /* Nothing really changed, but let's issue an notification call nonetheless,
+                                 * in case somebody is waiting for this. (e.g. file system ro/rw
+                                 * remounts.) */
                                 mount_set_state(mount, mount->state);
                                 break;
                         }