diff --git a/libs/image/kis_update_job_item.h b/libs/image/kis_update_job_item.h
--- a/libs/image/kis_update_job_item.h
+++ b/libs/image/kis_update_job_item.h
@@ -70,8 +70,8 @@
 
         setDone();
 
-        emit sigDoSomeUsefulWork();
         emit sigJobFinished();
+        emit sigDoSomeUsefulWork();
 
         m_exclusiveJobLock->unlock();
     }
diff --git a/libs/image/kis_update_scheduler.cpp b/libs/image/kis_update_scheduler.cpp
--- a/libs/image/kis_update_scheduler.cpp
+++ b/libs/image/kis_update_scheduler.cpp
@@ -55,15 +55,19 @@
 
     KisSimpleUpdateQueue updatesQueue;
     KisStrokesQueue strokesQueue;
-    KisUpdaterContext updaterContext;
     bool processingBlocked = false;
     qreal balancingRatio = 1.0; // updates-queue-size/strokes-queue-size
     KisProjectionUpdateListener *projectionUpdateListener;
     KisQueuesProgressUpdater *progressUpdater = 0;
 
     QAtomicInt updatesLockCounter;
     QReadWriteLock updatesStartLock;
     KisLazyWaitCondition updatesFinishedCondition;
+
+    // KisUpdaterContext can emit signals to KisUpdateScheduler in the dtor, so it
+    // must to be deleted before anything else.
+    // That means updaterContext must be declared last.
+    KisUpdaterContext updaterContext;
 };
 
 KisUpdateScheduler::KisUpdateScheduler(KisProjectionUpdateListener *projectionUpdateListener)
diff --git a/libs/image/kis_updater_context.h b/libs/image/kis_updater_context.h
--- a/libs/image/kis_updater_context.h
+++ b/libs/image/kis_updater_context.h
@@ -23,6 +23,7 @@
 #include <QMutex>
 #include <QReadWriteLock>
 #include <QThreadPool>
+#include <QSemaphore>
 
 #include "kis_base_rects_walker.h"
 #include "kis_async_merger.h"
@@ -128,7 +129,11 @@
 protected:
     static bool walkerIntersectsJob(KisBaseRectsWalkerSP walker,
                                     const KisUpdateJobItem* job);
+
+    qint32 defaultThreadCount() const;
+
     qint32 findSpareThread();
+    void startJob(KisUpdateJobItem* job);
 
 protected:
     /**
@@ -141,6 +146,7 @@
 
     QMutex m_lock;
     QVector<KisUpdateJobItem*> m_jobs;
+    QSemaphore m_waitSema;
     QThreadPool m_threadPool;
     KisLockFreeLodCounter m_lodCounter;
 };
diff --git a/libs/image/kis_updater_context.cpp b/libs/image/kis_updater_context.cpp
--- a/libs/image/kis_updater_context.cpp
+++ b/libs/image/kis_updater_context.cpp
@@ -25,14 +25,10 @@
 #include "kis_stroke_job.h"
 
 
-KisUpdaterContext::KisUpdaterContext(qint32 threadCount)
+KisUpdaterContext::KisUpdaterContext(qint32 threadCount):
+    m_jobs(threadCount > 0 ? threadCount : defaultThreadCount()),
+    m_waitSema(m_jobs.size())
 {
-    if(threadCount <= 0) {
-        threadCount = QThread::idealThreadCount();
-        threadCount = threadCount > 0 ? threadCount : 1;
-    }
-
-    m_jobs.resize(threadCount);
     for(qint32 i = 0; i < m_jobs.size(); i++) {
         m_jobs[i] = new KisUpdateJobItem(&m_exclusiveJobLock);
         connect(m_jobs[i], SIGNAL(sigContinueUpdate(const QRect&)),
@@ -54,6 +50,12 @@
         delete m_jobs[i];
 }
 
+qint32 KisUpdaterContext::defaultThreadCount() const
+{
+    int threadCount = QThread::idealThreadCount();
+    return threadCount > 0 ? threadCount : 1;
+}
+
 void KisUpdaterContext::getJobsSnapshot(qint32 &numMergeJobs,
                                         qint32 &numStrokeJobs)
 {
@@ -121,7 +123,7 @@
     Q_ASSERT(jobIndex >= 0);
 
     m_jobs[jobIndex]->setWalker(walker);
-    m_threadPool.start(m_jobs[jobIndex]);
+    startJob(m_jobs[jobIndex]);
 }
 
 /**
@@ -144,7 +146,7 @@
     Q_ASSERT(jobIndex >= 0);
 
     m_jobs[jobIndex]->setStrokeJob(strokeJob);
-    m_threadPool.start(m_jobs[jobIndex]);
+    startJob(m_jobs[jobIndex]);
 }
 
 /**
@@ -167,7 +169,7 @@
     Q_ASSERT(jobIndex >= 0);
 
     m_jobs[jobIndex]->setSpontaneousJob(spontaneousJob);
-    m_threadPool.start(m_jobs[jobIndex]);
+    startJob(m_jobs[jobIndex]);
 }
 
 /**
@@ -185,7 +187,9 @@
 
 void KisUpdaterContext::waitForDone()
 {
-    m_threadPool.waitForDone();
+    qint32 allJobs = m_jobs.size();
+    m_waitSema.acquire(allJobs);
+    m_waitSema.release(allJobs);
 }
 
 bool KisUpdaterContext::walkerIntersectsJob(KisBaseRectsWalkerSP walker,
@@ -204,9 +208,17 @@
     return -1;
 }
 
+
+void KisUpdaterContext::startJob(KisUpdateJobItem* job)
+{
+    m_waitSema.acquire();
+    m_threadPool.start(job);
+}
+
 void KisUpdaterContext::slotJobFinished()
 {
     m_lodCounter.removeLod();
+    m_waitSema.release();
 
     // Be careful. This slot can be called asynchronously without locks.
     emit sigSpareThreadAppeared();