Index: runs.scm
==================================================================
--- runs.scm
+++ runs.scm
@@ -16,11 +16,12 @@
 ;;     along with Megatest.  If not, see <http://www.gnu.org/licenses/>.
 
 ;;  strftime('%m/%d/%Y %H:%M:%S','now','localtime')
 
 (use (prefix sqlite3 sqlite3:) srfi-1 posix regex regex-case srfi-69 (srfi 18) 
-     posix-extras directory-utils pathname-expand typed-records format  sxml-serializer sxml-modifications)
+     posix-extras directory-utils pathname-expand typed-records format  sxml-serializer
+     sxml-modifications matchable)
 
 (declare (unit runs))
 (declare (uses db))
 (declare (uses common))
 (declare (uses items))
@@ -1593,19 +1594,10 @@
 			   itemmaps:    itemmaps
 			   ;; prereqs-not-met: prereqs-not-met
 			   )))
 	(runs:dat-regfull-set! runsdat regfull)
     
-        ;; -- removed BB 17ww28 - no longer needed.
-	;; every 15 minutes verify the server is there for this run
-	;; (if (and (common:low-noise-print 240 "try start server"  run-id)
-	;; 	 (not (or (and *runremote*
-	;; 		       (remote-server-url *runremote*)
-	;; 		       (server:ping (remote-server-url *runremote*)))
-	;; 		  (server:check-if-running *toppath*))))
-	;;     (server:kind-run *toppath*))
-	
 	(if (> num-running 0)
             (set! last-time-some-running (current-seconds)))
 
         (if (> (current-seconds)(+ last-time-some-running (or (configf:lookup *configdat* "setup" "give-up-waiting") 36000)))
             (hash-table-set! *max-tries-hash* tfullname (+ (hash-table-ref/default *max-tries-hash* tfullname 0) 1)))
@@ -1648,11 +1640,11 @@
 		     "\n  regfull:     " regfull
 		     "\n  reglen:      " reglen
 		     "\n  length reg:  " (length reg)
                      )
 
-	(runs:parallel-runners-mgmt runsdat)
+	;; (runs:parallel-runners-mgmt runsdat)
 
 	;; check for hed in waitons => this would be circular, remove it and issue an
 	;; error
 	(if (member test-name waitons)
 	    (begin
@@ -1688,14 +1680,35 @@
 	      (loop (car tal)(cdr tal) reg reruns))
 	  (runs:testdat-prereqs-not-met-set! testdat (rmt:get-prereqs-not-met run-id waitons hed item-path mode: testmode itemmaps: itemmaps))
 
 	  ;; This would be a good place to block on homehost load
 
-
-
-	  
-	  (runs:dat-can-run-more-tests-set! runsdat (runs:can-run-more-tests runsdat run-id jobgroup max-concurrent-jobs)) ;; look at the test jobgroup and tot jobs running
+	  ;; gonna try a strategy change here.
+	  ;;
+	  ;; check if can run more tests. if yes, continue, if no, rest for 10 seconds, check again
+	  ;; repeat until can run more tests
+	  ;;
+	  ;; look at the test jobgroup and tot jobs running
+	  (let loop-can-run-more
+	      ((res      (runs:can-run-more-tests runsdat run-id jobgroup max-concurrent-jobs))
+	       (remtries 60))
+	    (match res
+              ((run-more num-running . rem)
+	       (if (or run-more
+		       (< remtries 1))
+		   (begin
+		     (if (runs:lownoise "num-running" 30)
+			 (debug:print-info 0 *default-log-port* "Have "num-running" tests of max " max-concurrent-jobs))
+		     (runs:dat-can-run-more-tests-set! runsdat res)) ;; capture the result and then drop through
+		   (begin
+		     (if (runs:lownoise "num-running" 10)
+			 (debug:print-info 0 *default-log-port* "Can't run more tests, have "num-running" tests of "
+					   max-concurrent-jobs " allowed."))
+		     (thread-sleep! 5) ;; if we've hit max concurrent jobs take a breather, nb// make this configurable
+		     (loop-can-run-more (runs:can-run-more-tests runsdat run-id jobgroup max-concurrent-jobs)
+					(- remtries 1)))))))
+	      
 	  (let ((loop-list (runs:process-expanded-tests runsdat testdat)))
             (if loop-list (apply loop loop-list))))
 
 	 ;; items processed into a list but not came in as a list been processed
 	 ;;