qtmux with non-seekable downstream collects all input buffers before outputting any
Since gst-plugins-good!732 (merged) if qtmux has a non-seekable downstream, it will in the default mode, collect the entire input stream into the output_buffers
list before outputting any of them.
Not only does that mean that memory usage is potentially unbounded, but also CPU consumption is dominated by g_list_append
walking the super-long output buffers list repeatedly.