Need clarification: Does the spec allow varying the video resolution sent to the sink per-frame?

For screen capturing use cases, such as desktop or tab capture, the source
content can be resized during a capture session (e.g., the user resizes
their browser window).  However, a video stream whose resolution changes
dynamically (i.e., possibly per-frame) at the sink is not discussed nor in
the examples.

>From what I can tell, the spec seems to always assume one fixed video
resolution is chosen at the beginning, within what is allowed by the
MediaTrackConstraints.  Then, only by using applyConstraints() can a
different fixed resolution be set (by evaluating the new constraints).

Or, am I wrong, and because it's not explicitly stated, that there is NO
assumption about video resolution to the sink being fixed?  For what it's
worth, in the Chrome implementation today, the resolution *can* vary
per-frame.

*Details on my use case:* In Chrome, our tab capture implementation uses
the min and max width and height constraints to produce video frames of any
resolution within range.  It can also provide a fixed aspect ratio, with
letter-boxing added where needed, or use any aspect ratio.

In addition, video resolution can change, not just because of source
content resizes, but also depending on system performance (CPU/GPU
availability) and other external conditions (e.g., network bandwidth when
remoting content).  For example, on a low-end machine, video resolution
will be decreased in order to maintain a smooth frame rate.  For further
details, please see:
https://www.chromium.org/developers/design-documents/auto-throttled-screen-capture-and-mirroring

Thanks,
Yuri