doug@s15:~/linux-3.0.0/kernel$ diff -u sched.c.original sched.c.doug05
--- sched.c.original    2012-02-09 11:38:59.376250335 -0800
+++ sched.c.doug05      2012-02-28 08:42:56.805450230 -0800
@@ -3258,6 +3258,7 @@
 /* Variables and functions for calc_load */
 static atomic_long_t calc_load_tasks;
 static unsigned long calc_load_update;
+static unsigned calc_load_update_done_already=0;
 unsigned long avenrun[3];
 EXPORT_SYMBOL(avenrun);

@@ -3403,14 +3404,19 @@
         * If we crossed a calc_load_update boundary, make sure to fold
         * any pending idle changes, the respective CPUs might have
         * missed the tick driven calc_load_account_active() update
-        * due to NO_HZ.
+        * due to NO_HZ. Handshake ensures it is only done once.
         */
-       delta = calc_load_fold_idle();
-       if (delta)
-               atomic_long_add(delta, &calc_load_tasks);
-
+       if(calc_load_update_done_already == 0){
+               delta = calc_load_fold_idle();
+               if (delta)
+                        atomic_long_add(delta, &calc_load_tasks);
+               calc_load_update_done_already = 1;
+       }
        /*
         * If we were idle for multiple load cycles, apply them.
+        * (delete this note in final patch) Note: this code is never executed
+        * on my server systems. The point being that somehow this path needs
+        * to be tested with the proposed patch, but I don't know how.
         */
        if (ticks >= LOAD_FREQ) {
                n = ticks / LOAD_FREQ;
@@ -3487,6 +3493,7 @@
        avenrun[2] = calc_load(avenrun[2], EXP_15, active);

        calc_load_update += LOAD_FREQ;
+       calc_load_update_done_already = 0;
 }

 /*