Accurate terminal state tracking
The terminal state capture method used for force refreshing, upon connections to live sessions, and upon playback from moments, is a rough and incomplete prototype implementation. For many terminal based workflows, the existing implementation will not present major problems. However, some programs that do large numbers of small updates to the screen will be visibly affected by the limitations imposed by the current state capture.
For historical playback, the best workaround is to begin playback at a moment further back in time, and use an accelerated playback rate to reach the time of interest.
The current "state tracker" for terminal contents is a dumb ring buffer, with read-only analyzers on the side for resizing/retitling events, and crude injection of title customization events. This implementation was chosen specifically to avoid implementing full terminal state tracking for initial experimentation, and the straight-through replay of raw data from the start of a recording would be guaranteed to produce correct output if sent into a terminal emulator.
This approach is not correct when replaying data from points after the very start of the recording. For experimental tests, it was "good enough" for conceptual validation, but it must be replaced with a full state tracker to avoid showing artifacts already demonstrated in the pre-alpha manual.
Full terminal state tracking would also permit overlaying additional visual interactive UI for certain tasks in both control and playback modes, as well as framing/multi-panelling of terminal events.