In tmux 3.7 there’s a new protocol that I’ll use to optimize long agent history redraw in Tau, by making them an atomic logical repain. So tmux doesn’t need to output the whole redrawn scrollback to the terminal at all.

## The better, explicit mechanism

**tmux 3.7+ supports application-initiated synchronized output (`DECSET 2026`).** An application c
an bracket one logical repaint:

```text
CSI ? 2026 h
...clear and redraw...
CSI ? 2026 l
```

tmux updates its internal state while suppressing intermediate cell output, then emits the final d
irty view. A one-second timeout protects against a broken application. This is almost exactly the
optimization you described, but the application must identify repaint boundaries.