diff --git a/mps/code/comm.gmk b/mps/code/comm.gmk
index 5c181c99790..655a4d6c060 100644
--- a/mps/code/comm.gmk
+++ b/mps/code/comm.gmk
@@ -346,10 +346,8 @@ testratio: phony
 MMQA=perl test/qa -p $(PFM) -v $(VARIETY)
 
 $(PFM)/$(VARIETY)/testmmqa:
-	(if [ "$(VARIETY)" = "cool" ]; then cd ../test && $(MMQA) runset testsets/coolonly; fi)
-	(cd ../test && $(MMQA) runset testsets/argerr)
-	(cd ../test && $(MMQA) runset testsets/conerr)
-	(cd ../test && $(MMQA) runset testsets/passing)
+	([ "$(VARIETY)" = "cool" ] && cd ../test && $(MMQA) runset testsets/coolonly)
+	(cd ../test && $(MMQA) runset testsets/{argerr,conerr,passing})
 
 
 # == Toy Scheme interpreter ==
diff --git a/mps/procedure/release-build.rst b/mps/procedure/release-build.rst
index 24149fab272..2485c6140a7 100644
--- a/mps/procedure/release-build.rst
+++ b/mps/procedure/release-build.rst
@@ -111,6 +111,8 @@ All relative paths are relative to
         nmake /f w3i6mv.nmk clean testci
         nmake /f ananmv.nmk clean testansi
         nmake /f ananmv.nmk CFLAGS="-DCONFIG_POLL_NONE" clean testpollnone
+        cd ../test
+        perl test/qa runset testsets/{coolonly,argerr,conerr,passing}
 
 #. Check that there are no performance regressions by comparing the
    benchmarks (``djbench`` and ``gcbench``) for the last release and
diff --git a/mps/test/test/script/commands/runset b/mps/test/test/script/commands/runset
index 21224fc2297..3e579262392 100644
--- a/mps/test/test/script/commands/runset
+++ b/mps/test/test/script/commands/runset
@@ -12,7 +12,9 @@
  "LOG_DIR"
 );
 
-$testset = $ARGV[0];
-
-&run_testset($testset, "$LOG_DIR/summ.log", "$LOG_DIR/res.log", "$LOG_DIR/full.log");
+@LOGS = ("$LOG_DIR/summ.log", "$LOG_DIR/res.log", "$LOG_DIR/full.log");
+unlink(@LOGS);
+foreach $testset (@ARGV) {
+ &run_testset($testset, @LOGS);
+}
 
diff --git a/mps/test/test/script/help/runset b/mps/test/test/script/help/runset
index 1e3c503ee37..2b2576b1179 100644
--- a/mps/test/test/script/help/runset
+++ b/mps/test/test/script/help/runset
@@ -1,10 +1,10 @@
-run tests from a test-set
+run tests from test-sets
 % $Id$
 
-Usage: qa runset [<options>] <testsetfile>
+Usage: qa runset [<options>] <testsetfile> [<testsetfile> ...]
 
 'runset' compiles the test libraries and then runs all the
-tests in a test-set.
+tests in the test-sets.
 
 A test-set is a file containing a list of tests, one test
 per line. Blank lines or those beginning with % are ignored.
diff --git a/mps/test/test/script/runtest b/mps/test/test/script/runtest
index 192e62d4f5c..a929b6cf4a3 100644
--- a/mps/test/test/script/runtest
+++ b/mps/test/test/script/runtest
@@ -236,9 +236,9 @@ sub run_testset {
 
  %testsetresults = ();
 
- open(LOG_SUMMARY, ">".$logsummfile);
- open(LOG_RESULTS, ">".$logresfile);
- open(LOG_FULL,    ">".$logfullfile);
+ open(LOG_SUMMARY, ">>".$logsummfile);
+ open(LOG_RESULTS, ">>".$logresfile);
+ open(LOG_FULL,    ">>".$logfullfile);
  &describe_settings(LOG_SUMMARY);
  @LOG_FILES = (STDOUT, LOG_SUMMARY, LOG_RESULTS, LOG_FULL);
  &logcomment("Test set $testsetfile");