A spherical canvas could represent a photo sphere, or maps of the Earth or other planet surfaces, or the stars around the earth. Unlike the standard euclidean infinite canvas, this would have finite area but retain infinite zoom depth. Useful nodes could, for example:

• Provide base maps for GIS data such as satellite/aerial imagery
• Digital elevation maps
• OpenStreetMap layers
• Import custom survey datasets and lat/long/(elevation?) tracks
• Generate star charts from star databases
• Plot tracks for satellites, planets, comets, etc. from their ephemera
• Visualize KML files

Things you could do:

• Make videos with motion graphics incorporating city street/transit layouts or city/country/continent-scale boundaries
• Annotating maps or overlaying CAD site plans on a map
• Visualize sea level rise or flooding
• Process LiDAR or other survey datasets with image processing techniques
• Generate celestial body tracks over star charts
• Animate a planetarium show (and render its different projector views)
• Use tools to paint, draw, and animate over photo spheres and 360° videos
• Load accelerometer data to stabilize and reorient the spherical canvas
• Render out tiny planet projections
• Design skyboxes for games/animations

Besides spherical coordinates, other non-euclidean canvas coordinate systems could be used, such as a finite tiling rectangle or hex cell. Drawing off one edge would result in drawing on the opposite edge. This would be more sophisticated than an infinite tiling node because the actual coordinate space would wrap, not just the content being generated infinitely. Other variants of wrapping could include a ping-ponging coordinate space or a Möbius strip (see this video).

See https://en.wikipedia.org/wiki/Template:Orthogonal_coordinate_systems and https://en.wikipedia.org/wiki/Orthogonal_coordinates

This all somewhat relates to #1271 which will involve moving beyond affine transforms.