Currently, only one AR frame is supported at a time. There are various tasks needed to support multiple frames properly including:

Some cleanup in xr_webgl_drawing_buffer.cc::OverwriteColorBufferFromMailboxTexture() and its usage:
It would be cleaner to do this calculation in the calling code and just pass in the final size.

if source < dest, should we clear, or are sites expected to clear themselves?
if source != dest, do we need to make sure to update projection matrices?  If so, it might be nice to pass in dest_size here and do that calculation in xr code so we know the right projection matrix.
use ShrunkTo - copy_size = size_in.ShrunkTo(size_) to replace the two std::min lines

Anchors need to be filtered by session so sessions can't see each other's anchors.
 

Currently, only one AR XRSession is supported at a time. There are various tasks needed to support multiple AR iframes/tabs properly including:

Some cleanup in xr_webgl_drawing_buffer.cc::OverwriteColorBufferFromMailboxTexture() and its usage:
It would be cleaner to do this calculation in the calling code and just pass in the final size.

if source < dest, should we clear, or are sites expected to clear themselves?
if source != dest, do we need to make sure to update projection matrices?  If so, it might be nice to pass in dest_size here and do that calculation in xr code so we know the right projection matrix.
use ShrunkTo - copy_size = size_in.ShrunkTo(size_) to replace the two std::min lines

Anchors need to be filtered by session so sessions can't see each other's anchors.