diff --git a/include/mimalloc/types.h b/include/mimalloc/types.h
index 5dfbb808..ba9a8864 100644
--- a/include/mimalloc/types.h
+++ b/include/mimalloc/types.h
@@ -300,7 +300,7 @@ typedef struct mi_page_s {
 // ------------------------------------------------------
 
 #define MI_PAGE_ALIGN                     MI_ARENA_SLICE_ALIGN // pages must be aligned on this for the page map.
-#define MI_PAGE_MIN_BLOCK_ALIGN           (64)                 // minimal block alignment in a page 
+#define MI_PAGE_MIN_BLOCK_ALIGN           MI_SIZE_BITS         // minimal block alignment in a page (64b on 64-bit, 32b on 32-bit)
 #define MI_PAGE_MAX_OVERALLOC_ALIGN       MI_ARENA_SLICE_SIZE  // (64 KiB) limit for which we overallocate in arena pages, beyond this use OS allocation
 
 #if (MI_ENCODE_FREELIST || MI_PADDING) && MI_SIZE_SIZE == 8
diff --git a/src/bitmap.c b/src/bitmap.c
index 2ef692cb..7f4c8776 100644
--- a/src/bitmap.c
+++ b/src/bitmap.c
@@ -805,10 +805,10 @@ static bool mi_bitmap_chunkmap_try_clear(mi_bitmap_t* bitmap, size_t chunk_idx)
     return false;
   }
   // record the max clear 
-  /*size_t oldmax = mi_atomic_load_relaxed(&bitmap->chunk_max_clear);
+  size_t oldmax = mi_atomic_load_relaxed(&bitmap->chunk_max_clear);
   do {
     if mi_likely(chunk_idx <= oldmax) break;
-  } while (!mi_atomic_cas_weak_acq_rel(&bitmap->chunk_max_clear, &oldmax, chunk_idx));*/
+  } while (!mi_atomic_cas_weak_acq_rel(&bitmap->chunk_max_clear, &oldmax, chunk_idx));
   return true;
 }
 
@@ -1042,21 +1042,23 @@ bool mi_bitmap_is_xsetN(mi_xset_t set, mi_bitmap_t* bitmap, size_t idx, size_t n
 -------------------------------------------------------------------------------- */
 
 
-#define mi_bitmap_forall_chunks(bitmap, tseq, name_epoch, name_chunk_idx) \
+#define mi_bitmap_forall_chunks(bitmap, tseq, name_chunk_idx) \
   { \
   /* start chunk index -- todo: can depend on the tseq to decrease contention between threads */ \
   MI_UNUSED(tseq); \
-  /* const size_t chunk_max = mi_atomic_load_acquire(&bitmap->chunk_max_clear); */ /* mi_bitmap_chunk_count(bitmap) */ \
-  const size_t chunk_start = 0; /* (chunk_max <= 1 ? 0 : (tseq % chunk_max)); */      /* space out threads */ \
+  const size_t chunk_start = 0; /* (tseq % (1+chunk_hi_idx)); */ /* space out threads? */ \
   const size_t chunkmap_max_bfield = _mi_divide_up( mi_bitmap_chunk_count(bitmap), MI_BCHUNK_BITS ); \
+  const size_t chunkmap_hi_bfield  = chunkmap_max_bfield; /* chunk_hi_idx / MI_BFIELD_BITS; */\
   const size_t chunkmap_start = chunk_start / MI_BFIELD_BITS; \
   const size_t chunkmap_start_idx = chunk_start % MI_BFIELD_BITS; \
   /* for each chunkmap entry `i` */ \
   for (size_t _i = 0; _i < chunkmap_max_bfield; _i++) { \
-    size_t i = (_i + chunkmap_start); \
-    if (i >= chunkmap_max_bfield) { \
-      i -= chunkmap_max_bfield; /* adjust for the start position */ \
+    size_t i; \
+    if (_i < chunkmap_hi_bfield) { \
+      i = _i + chunkmap_start; /* first the chunks up to chunk_hi */ \
+      if (i >= chunkmap_hi_bfield) { i -= chunkmap_hi_bfield; } /* rotate */ \
     } \
+    else { i = _i;  }  /* the rest of the chunks above chunk_hi_idx */ \
     const size_t chunk_idx0 = i*MI_BFIELD_BITS; \
     mi_bfield_t cmap = mi_atomic_load_relaxed(&bitmap->chunkmap.bfields[i]); \
     size_t      cmap_idx_shift = 0;   /* shift through the cmap */ \
@@ -1086,7 +1088,8 @@ bool mi_bitmap_is_xsetN(mi_xset_t set, mi_bitmap_t* bitmap, size_t idx, size_t n
 // (Used to find fresh free slices.)
 mi_decl_nodiscard bool mi_bitmap_try_find_and_clearN(mi_bitmap_t* bitmap, size_t n, size_t tseq, size_t* pidx)
 {
-  mi_bitmap_forall_chunks(bitmap, tseq, epoch, chunk_idx)
+  // const size_t chunk_hi_idx = mi_atomic_load_relaxed(&bitmap->chunk_max_clear);
+  mi_bitmap_forall_chunks(bitmap, tseq, chunk_idx)
   {
     size_t cidx;
     if mi_likely(mi_bchunk_find_and_try_clearN(&bitmap->chunks[chunk_idx], n, &cidx)) {
@@ -1111,7 +1114,7 @@ mi_decl_nodiscard bool mi_bitmap_try_find_and_clearN(mi_bitmap_t* bitmap, size_t
 mi_decl_nodiscard bool mi_bitmap_try_find_and_claim(mi_bitmap_t* bitmap, size_t tseq, size_t* pidx,
                                                     mi_claim_fun_t* claim, void* arg1, void* arg2)
 {
-  mi_bitmap_forall_chunks(bitmap, tseq, epoch, chunk_idx)
+  mi_bitmap_forall_chunks(bitmap, tseq, chunk_idx)
   {
     size_t cidx;
     if mi_likely(mi_bchunk_find_and_try_clear(&bitmap->chunks[chunk_idx], &cidx)) {
diff --git a/src/page-map.c b/src/page-map.c
index ca0e2481..d849e6a2 100644
--- a/src/page-map.c
+++ b/src/page-map.c
@@ -43,9 +43,9 @@ bool _mi_page_map_init(void) {
     bool is_zero;
     _mi_os_commit(_mi_page_map, _mi_os_page_size(), &is_zero, NULL);
     if (!is_zero && !mi_page_map_memid.initially_zero) { _mi_memzero(_mi_page_map, _mi_os_page_size()); }
-    _mi_page_map[0] = 1; // so _mi_ptr_page(NULL) == NULL
-    mi_assert_internal(_mi_ptr_page(NULL)==NULL);
   }
+  _mi_page_map[0] = 1; // so _mi_ptr_page(NULL) == NULL
+  mi_assert_internal(_mi_ptr_page(NULL)==NULL);
   return true;
 }
 
diff --git a/test/main-override.cpp b/test/main-override.cpp
index 3f64117a..5a1fc6d2 100644
--- a/test/main-override.cpp
+++ b/test/main-override.cpp
@@ -382,7 +382,7 @@ static void test_mt_shutdown()
 
 // issue #372
 static void fail_aslr() {
-  size_t sz = (4ULL << 40); // 4TiB
+  uint64_t sz = (4ULL << 40); // 4TiB
   void* p = malloc(sz);
   printf("pointer p: %p: area up to %p\n", p, (uint8_t*)p + sz);
   *(int*)0x5FFFFFFF000 = 0;  // should segfault
diff --git a/test/test-stress.c b/test/test-stress.c
index d46c2484..19edf2b5 100644
--- a/test/test-stress.c
+++ b/test/test-stress.c
@@ -40,7 +40,7 @@ static int ITER    = 20;
 static int THREADS = 8;
 static int SCALE   = 10;
 static int ITER    = 10;
-#elif 1
+#elif 0
 static int THREADS = 4;
 static int SCALE   = 100;
 static int ITER    = 10;