[Bug 51219] Deferring EC_COMPLETE introduced two new quartz:filtergraph failures

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51219

--- Comment #5 from Anton Baskanov  ---
The failing test case ("test.mpg") is a bit unusual as EOS happens before the
first frame.

IIRC the issue is that the streaming thread gets stuck trying to enter the
filter CS while sending EOS. It happens because async_run_cb() calls
filter_GetState() in a loop which reenters the filter CS immediately after it
was released. It is somewhat similar to bug 53403.

The function that locks the filter CS on the streaming thread is
pin_ConnectedTo():
sink_EndOfStream() -> sink_eos() -> strmbase_passthrough_eos() ->
MediaSeekingPassThru_GetStopPosition() -> get_connected() -> pin_ConnectedTo().