Belts carry 4 grains deep; stalled belts warn and stop animating (t45)

docs/TESTING.md

@@ -14,6 +14,19 @@ prints the failing checks.
 
 ## Headless suites
 
+For t45, `node test/t_stalls.js` checks the four-grain bed, sustained local jams,
+recovery, bars, cleated runs, slower screens and an overloaded closed loop.
+`test/t_threads.js` includes loaded belts and slopes, and hashes stall/animation state
+as well as cell bytes with inline, 1, 2 and 8 workers (including repeats).
+`python3 test/t_stalls.py` checks both looks: hover advice, always-visible warnings,
+fog, actual frozen belt/filter/screw pixels and clearing the blockage. It saves
+`shots/t45-stalled.png` and `shots/t45-stalled-smooth.png`.
+`node tools/experiments/belt-depth.js` measures sustained one- and sixteen-layer feeds
+on a flat belt: seed 7, 220 warm-up ticks and 1200 measured ticks, with an open sink.
+The side-fed lift comparison in `t_conveyors.js` now uses a four-grain bed in both
+fixtures rather than the old sixty-grain heap and its hard-coded delivery baseline;
+it retains the 90% comparison, conservation and all-lane checks.
+
 For t69, `node test/t_screw_heat.js` checks full-cost repair, load preservation, undo,
 all screw directions, the unchanged melt threshold, and saved damage/message state.
 `python3 test/t_screw_heat.py` is the focused browser check for both looks, hover,