diff --git a/compiler-rt/lib/scudo/standalone/combined.h b/compiler-rt/lib/scudo/standalone/combined.h
--- a/compiler-rt/lib/scudo/standalone/combined.h
+++ b/compiler-rt/lib/scudo/standalone/combined.h
@@ -588,8 +588,9 @@
     return Options.MayReturnNull;
   }
 
-  // TODO(kostyak): implement this as a "backend" to mallopt.
-  bool setOption(UNUSED uptr Option, UNUSED uptr Value) { return false; }
+  void setReleaseToOsIntervalMs(s32 Interval) {
+    Primary.setReleaseToOsIntervalMs(Interval);
+  }
 
   // Return the usable size for a given chunk. Technically we lie, as we just
   // report the actual size of a chunk. This is done to counteract code actively
diff --git a/compiler-rt/lib/scudo/standalone/primary32.h b/compiler-rt/lib/scudo/standalone/primary32.h
--- a/compiler-rt/lib/scudo/standalone/primary32.h
+++ b/compiler-rt/lib/scudo/standalone/primary32.h
@@ -73,11 +73,10 @@
       SizeClassInfo *Sci = getSizeClassInfo(I);
       Sci->RandState = getRandomU32(&Seed);
       // See comment in the 64-bit primary about releasing smaller size classes.
-      Sci->CanRelease = (ReleaseToOsInterval >= 0) &&
-                        (I != SizeClassMap::BatchClassId) &&
+      Sci->CanRelease = (I != SizeClassMap::BatchClassId) &&
                         (getSizeByClassId(I) >= (PageSize / 32));
     }
-    ReleaseToOsIntervalMs = ReleaseToOsInterval;
+    setReleaseToOsIntervalMs(ReleaseToOsInterval);
   }
   void init(s32 ReleaseToOsInterval) {
     memset(this, 0, sizeof(*this));
@@ -175,6 +174,10 @@
       getStats(Str, I, 0);
   }
 
+  void setReleaseToOsIntervalMs(s32 Interval) {
+    atomic_store(&ReleaseToOsIntervalMs, Interval, memory_order_release);
+  }
+
   uptr releaseToOS() {
     uptr TotalReleasedBytes = 0;
     for (uptr I = 0; I < NumClasses; I++) {
@@ -363,6 +366,10 @@
                 AvailableChunks, Rss >> 10);
   }
 
+  s32 getReleaseToOsIntervalMs() {
+    return atomic_load(&ReleaseToOsIntervalMs, memory_order_acquire);
+  }
+
   NOINLINE uptr releaseToOSMaybe(SizeClassInfo *Sci, uptr ClassId,
                                  bool Force = false) {
     const uptr BlockSize = getSizeByClassId(ClassId);
@@ -381,7 +388,7 @@
     }
 
     if (!Force) {
-      const s32 IntervalMs = ReleaseToOsIntervalMs;
+      const s32 IntervalMs = getReleaseToOsIntervalMs();
       if (IntervalMs < 0)
         return 0;
       if (Sci->ReleaseInfo.LastReleaseAtNs +
@@ -419,7 +426,7 @@
   // through the whole NumRegions.
   uptr MinRegionIndex;
   uptr MaxRegionIndex;
-  s32 ReleaseToOsIntervalMs;
+  atomic_s32 ReleaseToOsIntervalMs;
   // Unless several threads request regions simultaneously from different size
   // classes, the stash rarely contains more than 1 entry.
   static constexpr uptr MaxStashedRegions = 4;
diff --git a/compiler-rt/lib/scudo/standalone/primary64.h b/compiler-rt/lib/scudo/standalone/primary64.h
--- a/compiler-rt/lib/scudo/standalone/primary64.h
+++ b/compiler-rt/lib/scudo/standalone/primary64.h
@@ -86,12 +86,11 @@
       // memory accesses which ends up being fairly costly. The current lower
       // limit is mostly arbitrary and based on empirical observations.
       // TODO(kostyak): make the lower limit a runtime option
-      Region->CanRelease = (ReleaseToOsInterval >= 0) &&
-                           (I != SizeClassMap::BatchClassId) &&
+      Region->CanRelease = (I != SizeClassMap::BatchClassId) &&
                            (getSizeByClassId(I) >= (PageSize / 32));
       Region->RandState = getRandomU32(&Seed);
     }
-    ReleaseToOsIntervalMs = ReleaseToOsInterval;
+    setReleaseToOsIntervalMs(ReleaseToOsInterval);
 
     if (SupportsMemoryTagging)
       UseMemoryTagging = systemSupportsMemoryTagging();
@@ -187,6 +186,10 @@
       getStats(Str, I, 0);
   }
 
+  void setReleaseToOsIntervalMs(s32 Interval) {
+    atomic_store(&ReleaseToOsIntervalMs, Interval, memory_order_release);
+  }
+
   uptr releaseToOS() {
     uptr TotalReleasedBytes = 0;
     for (uptr I = 0; I < NumClasses; I++) {
@@ -244,7 +247,7 @@
   uptr PrimaryBase;
   RegionInfo *RegionInfoArray;
   MapPlatformData Data;
-  s32 ReleaseToOsIntervalMs;
+  atomic_s32 ReleaseToOsIntervalMs;
   bool UseMemoryTagging;
 
   RegionInfo *getRegionInfo(uptr ClassId) const {
@@ -378,6 +381,10 @@
                 getRegionBaseByClassId(ClassId));
   }
 
+  s32 getReleaseToOsIntervalMs() {
+    return atomic_load(&ReleaseToOsIntervalMs, memory_order_acquire);
+  }
+
   NOINLINE uptr releaseToOSMaybe(RegionInfo *Region, uptr ClassId,
                                  bool Force = false) {
     const uptr BlockSize = getSizeByClassId(ClassId);
@@ -397,7 +404,7 @@
     }
 
     if (!Force) {
-      const s32 IntervalMs = ReleaseToOsIntervalMs;
+      const s32 IntervalMs = getReleaseToOsIntervalMs();
       if (IntervalMs < 0)
         return 0;
       if (Region->ReleaseInfo.LastReleaseAtNs +
diff --git a/compiler-rt/lib/scudo/standalone/wrappers_c.inc b/compiler-rt/lib/scudo/standalone/wrappers_c.inc
--- a/compiler-rt/lib/scudo/standalone/wrappers_c.inc
+++ b/compiler-rt/lib/scudo/standalone/wrappers_c.inc
@@ -154,9 +154,21 @@
                  SCUDO_PREFIX(malloc_enable));
 }
 
-INTERFACE WEAK int SCUDO_PREFIX(mallopt)(int param, UNUSED int value) {
+INTERFACE WEAK int SCUDO_PREFIX(mallopt)(int param, int value) {
   if (param == M_DECAY_TIME) {
-    // TODO(kostyak): set release_to_os_interval_ms accordingly.
+    if (SCUDO_ANDROID) {
+      switch (value) {
+        case 0:
+          value = 500;
+          break;
+        case 1:
+          value = 5000;
+          break;
+        default:
+          scudo::Printf("Scudo WARNING: invalid decay value (%d):\n", value);
+      }
+    }
+    SCUDO_ALLOCATOR.setReleaseToOsIntervalMs(value);
     return 1;
   } else if (param == M_PURGE) {
     SCUDO_ALLOCATOR.releaseToOS();