diff --git a/libs/ardour/session_click.cc b/libs/ardour/session_click.cc
index 87444686a3..557e1c1d12 100644
--- a/libs/ardour/session_click.cc
+++ b/libs/ardour/session_click.cc
@@ -58,8 +58,6 @@ Session::add_click (samplepos_t pos, bool emphasis)
 void
 Session::click (samplepos_t cycle_start, samplecnt_t nframes)
 {
-	vector<TempoMap::BBTPoint> points; // TODO use mempool allocator
-
 	if (_click_io == 0) {
 		return;
 	}
@@ -87,32 +85,64 @@ Session::click (samplepos_t cycle_start, samplecnt_t nframes)
 
 	/* range to check for clicks */
 	samplepos_t start = cycle_start + offset;
-	const samplepos_t end = start + nframes;
 	/* correct start, potentially */
 	start = max (start, (samplepos_t) 0);
 
-	if (end > start) {
-		_tempo_map->get_grid (points, start, end);
-	}
+	samplecnt_t remain = nframes;
 
-	if (distance (points.begin(), points.end()) == 0) {
-		goto run_clicks;
-	}
+	while (remain > 0) {
+		samplecnt_t move = remain;
 
-	for (vector<TempoMap::BBTPoint>::iterator i = points.begin(); i != points.end(); ++i) {
-		switch ((*i).beat) {
-		case 1:
-			add_click ((*i).sample, true);
-			break;
-		default:
-			if (click_emphasis_data == 0 || (Config->get_use_click_emphasis () == false) || (click_emphasis_data && (*i).beat != 1)) { // XXX why is this check needed ??  (*i).beat !=1 must be true here
-				add_click ((*i).sample, false);
+		Location* loop_location = locations()->auto_loop_location ();
+		if (loop_location) {
+			const samplepos_t loop_start = loop_location->start ();
+			const samplepos_t loop_end = loop_location->end ();
+			if (start >= loop_end) {
+				samplecnt_t off = (start - loop_end) % (loop_end - loop_start);
+				start = loop_start + off;
+				move = std::min (remain, loop_end - start);
+			} else if (start + move >= loop_end) {
+				move = std::min (remain, loop_end - start);
+			}
+			if (move == 0) {
+				start = loop_start;
+				const samplecnt_t looplen = loop_end - loop_start;
+				move = std::min (remain, looplen);
 			}
-			break;
 		}
+
+		const samplepos_t end = start + move;
+
+		vector<TempoMap::BBTPoint> points; // TODO use mempool allocator
+		_tempo_map->get_grid (points, start, end);
+
+		if (distance (points.begin(), points.end()) == 0) {
+			start += move;
+			remain -= move;
+			continue;
+		}
+
+		for (vector<TempoMap::BBTPoint>::iterator i = points.begin(); i != points.end(); ++i) {
+			if ((*i).sample < start || (*i).sample >= end) {
+				// FIXME: TempoMap::get_grid() returns duplicates and out of range points
+				continue;
+			}
+			switch ((*i).beat) {
+				case 1:
+					add_click ((*i).sample, true);
+					break;
+				default:
+					if (click_emphasis_data == 0 || (Config->get_use_click_emphasis () == false) || (click_emphasis_data && (*i).beat != 1)) { // XXX why is this check needed ??  (*i).beat !=1 must be true here
+						add_click ((*i).sample, false);
+					}
+					break;
+			}
+		}
+
+		start += move;
+		remain -= move;
 	}
 
-  run_clicks:
 	clickm.release ();
 	run_click (cycle_start, nframes);
 }
@@ -135,31 +165,59 @@ Session::run_click (samplepos_t start, samplepos_t nframes)
 	buf = bufs.get_audio(0).data();
 	memset (buf, 0, sizeof (Sample) * nframes);
 
+	/* given a large output latency, `start' can be offset by by > 1 cycle.
+	 * and needs to be mapped back into the loop-range */
+	Location* loop_location = locations()->auto_loop_location ();
+	bool crossloop = false;
+	samplecnt_t span = nframes;
+	if (loop_location) {
+		const samplepos_t loop_start = loop_location->start ();
+		const samplepos_t loop_end = loop_location->end ();
+		if (start >= loop_end) {
+			samplecnt_t off = (start - loop_end) % (loop_end - loop_start);
+			start = loop_start + off;
+			span = std::min (nframes, loop_end - start);
+		} else if (start + nframes >= loop_end) {
+			crossloop = true;
+			span = std::min (nframes, loop_end - start);
+		}
+	}
+
 	for (list<Click*>::iterator i = clicks.begin(); i != clicks.end(); ) {
 
-		samplecnt_t copy;
+		Click *clk = *i;
+
+		if (loop_location) {
+			const samplepos_t loop_start = loop_location->start ();
+			const samplepos_t loop_end = loop_location->end ();
+			/* remove any clicks that are outside loop location, and not currently playing */
+			if ((clk->start < loop_start || clk->start >= loop_end) && clk->offset == 0) {
+				delete clk;
+				i = clicks.erase (i);
+				continue;
+			}
+		}
+
 		samplecnt_t internal_offset;
-		Click *clk;
 
-		clk = *i;
-
-		if (clk->start < start) {
+		if (clk->start <= start || clk->offset > 0) {
 			internal_offset = 0;
-		} else {
+		} else if (clk->start < start + span) {
+			/* queue click at offset in current cycle */
 			internal_offset = clk->start - start;
+		} else if (crossloop) {
+			/* When loop wraps around in current cycle, take
+			 * clicks at loop-start into account */
+			const samplepos_t loop_start = loop_location->start ();
+			internal_offset = clk->start - loop_start + span;
 		}
 
-		if (nframes < internal_offset) {
-		         /* we've just located or something..
-			    effectively going backwards.
-			    lets get the flock out of here */
-		        break;
+		if (internal_offset >= nframes) {
+			break;
 		}
 
-		copy = min (clk->duration - clk->offset, nframes - internal_offset);
-
+		samplecnt_t copy = min (clk->duration - clk->offset, nframes - internal_offset);
 		memcpy (buf + internal_offset, &clk->data[clk->offset], copy * sizeof (Sample));
-
 		clk->offset += copy;
 
 		if (clk->offset >= clk->duration) {