Please describe the problem:
I can run pipeline with alsasink only one time. After unref-ing the pipeline, and creating new one with alsasink the new pipeline won't go into PLAYING state. ringbuffer hangs in opening device.

On the other hand, works fine with OSSsink or previous version (gst-plugins-base 0.10.11)

Steps to reproduce:
Testcase: (http://timon.stabilis.ru/~timon/alstest.c)
#include <gst/gst.h>

#define GST_CREATE_EL(el, kind, name) do {				\
	el = gst_element_factory_make(kind, name);			\
	g_assert(el);										\
	g_print("Created %s of kind %s with name %s, %p\n",	\
		#el, kind, name, el);							\
	if (el != pipe)										\
		gst_bin_add(GST_BIN(pipe), el);					\
} while (0)
#define SINK "alsa"

GMainLoop *loop;
GstElement *pipe, *src, *conv, *res, *sink;
int timer = 0;

void
CreatePipeline()
{
	GST_CREATE_EL(pipe, "pipeline", "pipeline");
	GST_CREATE_EL(src, "audiotestsrc", "src");
	GST_CREATE_EL(conv, "audioconvert", "converter");
	GST_CREATE_EL(res, "audioresample", "resampler");
	GST_CREATE_EL(sink, SINK"sink", sink);
	gst_element_link_many(src, conv, res, sink, NULL);
}

gboolean
IdleTimer(gpointer data)
{
	timer ++;
	g_print("Timer: %d\n", timer);
	if ((timer % 10) == 0) {
		gst_element_set_state(pipe, GST_STATE_NULL);
		g_main_loop_quit(loop);
	}
	return TRUE;
}

int
main(int ac, char **av)
{
	gst_init(&ac, &av);
	loop = g_main_loop_new(NULL, FALSE);
	g_timeout_add(100, IdleTimer, NULL);

	CreatePipeline();
	gst_element_set_state(pipe, GST_STATE_PLAYING);
	g_main_loop_run(loop);
	g_object_unref(pipe);

	g_print("done\n");

	CreatePipeline();
	gst_element_set_state(pipe, GST_STATE_PLAYING);
	g_main_loop_run(loop);
	g_object_unref(pipe);

	return (0);
}


Actual results:
First time the pipeline runs well.
Second time the pipeline doesn't change to PLAYING state, with the last message:
0:00:02.348000000 10497 0x502010 DEBUG           ringbuffer gstringbuffer.c:443:gst_ring_buffer_open_device:<audiosinkringbuffer1> opening device.
Have to kill program using ^C.

Expected results:
Hear sample tone, then gap, then sample tone again.

Does this happen every time?
YES.

Other information:
Test program: http://timon.stabilis.ru/~timon/alstest.c
results of test run with --gst-debug=5 http://timon.stabilis.ru/~timon/alstest-cvs.log.gz