Improving the performance of GLR based applications depends primarily on reducing network bandwidth. Also, more investigation is needed of how multiple concurrent GLR applications interact and what GLR server and client policies can best reduce contention and the time spent waiting for a rendering interval to be scheduled.
The rendering network bandwidth (the OpenGL commands sent to render the scene) can be managed through careful use of display lists, texture objects, and state set-up outside of render intervals using glrEstablishRenderState. This implies that GLR's performance is largely limited by the network bandwidth to read back images.