Warp Stabilize without having to nest or match sequence dimensions

I understand the wish, and how useful it would be!

But ... I think that would be ... difficult, if not currently impossible, to code.

Too many users seem to conceptualize Warp as working like say in-camera lens or sensor stabilization ... and it's not anything like those function at all. Those don't use any computational processes of image data, they physically move either lens elements or the sensor itself. Or both. So that the captured image is already 'fixed' as it is captured.

No computational load is needed by the camera in processing the image itself.

Warp works by analyzing per pixel the image data, to detect "motion", and then tries to computer how to best remove evidence of motion. So it will 'zoom' in on the image, recreate a few pixels at times, and twist and turn the image. It's an incredibly complicated processing process.

And is totally dependent on framesize of both the image and the sequence. 

Further, it is already the single, dramatically biggest computational effect in the app, even more so than Neat video noise remover. 

What you're asking would even require a vastly more complex, complicated, and processor-intensive code. As in ... yowza!