If I understand the code correctly, OpenSSH only injects delays and chaff when you're actively typing data through an interactive session directly to the ssh client itself (i.e. it's reading your key strokes from the TTY, as opposed to keystrokes being sent as opaque data over the X11 channel). There would be no disruption to X11 forwarding while you're not actively typing, and even then there's no delay if it can bundle key presses with non-interactive data. See https://github.com/openssh/openssh-portable/blob/088ca00/cli...
For a full desktop session, X11 over TCP should be much faster just because each X11 client app has it's own TCP connection, whereas for SSH X11 forwarding every app is tunneled over a single TCP connection.
> For a full desktop session, X11 over TCP should be much faster just because each X11 client app has it's own TCP connection, whereas for SSH X11 forwarding every app is tunneled over a single TCP connection.
But even with a single X11 connection, the difference in performance is huge between it being tunneled through SSH and not.
Also, whether you have one connection or multiple between 2 hosts, it's still the same bandwidth for the most part.