The test started crashing since:
r260383 CRASH (Expected: FAIL)
Likely due to r260380 (MediaPlayerPrivateMediaStreamAVFObjC should start play a newly added audio track if it is playing).
Thread 1 (Thread 0x7faa687b09c0 (LWP 95220)):
#0 _g_log_abort () at ../../Source/glib-2.58.1/glib/gmessages.c:557
#1 0x00007faa6aecf559 in g_logv () at ../../Source/glib-2.58.1/glib/gmessages.c:1371
#2 0x00007faa6aecf712 in g_log () at ../../Source/glib-2.58.1/glib/gmessages.c:1413
#3 0x00007faa6afadc53 in g_object_unref () at ../../Source/glib-2.58.1/gobject/gobject.c:3309
#4 g_object_unref () at ../../Source/glib-2.58.1/gobject/gobject.c:3238
#5 0x00007faa6cff569a in _gst_message_free () at ../../Source/gstreamer-1.16.1/gst/gstmessage.c:212
#6 0x00007faa6cfce317 in gst_message_unref () at ../../Source/gstreamer-1.16.1/gst/gstmessage.h:382
#7 gst_bus_source_dispatch () at ../../Source/gstreamer-1.16.1/gst/gstbus.c:840
#8 0x00007faa6aec86b8 in g_main_dispatch () at ../../Source/glib-2.58.1/glib/gmain.c:3182
#9 g_main_context_dispatch () at ../../Source/glib-2.58.1/glib/gmain.c:3847
#10 0x00007faa6aec8a78 in g_main_context_iterate () at ../../Source/glib-2.58.1/glib/gmain.c:3920
#11 0x00007faa6aec8d62 in g_main_loop_run () at ../../Source/glib-2.58.1/glib/gmain.c:4116
#13 0x00007faa7355a34d in int WebKit::AuxiliaryProcessMain<WebKit::WebProcess, WebKit::WebProcessMainGtk>(int, char**) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-4.0.so.37
#14 0x00007faa6a27809b in __libc_start_main (main=0x5581020a4c10 <main>, argc=4, argv=0x7fffc626ebf8, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7fffc626ebe8) at ../csu/libc-start.c:308
#15 0x00005581020a4c8a in _start ()
Bisecting, I got the crashes starting right before, in r260379 (bug210494) (tracked in the same build that started crashing).
As a result of fixing bug fast/mediastream/MediaStream-removeTrack-while-playing.html begins to crash in debug mode in an assertion. When a track is removed the observer is not removed and when it dies there are null objects in the observer weak hash set. Still, if you do a quick patch to remove the observer, things can still crash as track removal is racy and in release there' other crash. All this needs to be taclked as a general approach to fix track removal.
This test also started to flaky timeout frequently in the release bots since r271526 (Memory pressure changes).
No longer crashing. Closing and adjusting expectations.