Opening a scanned location used to involve several rough edges: a blank flash when the loader disappeared, the camera sometimes framing the wrong part of the model on first load, and a crash if you switched organizations before the viewer had a chance to initialize.
All of those are fixed. The loading overlay now stays visible until the point cloud is fully framed in the viewport — so when it clears, you see your scan in context, not a white screen. The camera is positioned correctly on first load, meaning the initial view shows the relevant geometry rather than an off-target angle. If you switch organizations and then open the scan tab, the viewer initializes properly instead of showing "Failed to load scan."
Switching between RGB and Segmented color modes is now instant. Previously, toggling modes triggered a fresh load of the point cloud data. Now the viewer switches the display mode in place, so the classification panel responds immediately without a reload.
Script loading for the viewer now runs in parallel phases rather than one after another. This cuts the number of sequential network round-trips roughly in half, which makes the viewer reach an interactive state faster — particularly noticeable on first open or after a cache clear.