diff --git a/subwave_nowplaying.star b/subwave_nowplaying.star
index 9f9e645..7abdbe1 100644
--- a/subwave_nowplaying.star
+++ b/subwave_nowplaying.star
@@ -368,14 +368,19 @@ GLITCH_STATES_PER_CYCLE = 4
 # height) -- see the note above GLITCH_HOLD_TICKS for why that's the only
 # part grounded in anything observed rather than invented.
 GLITCH_STATE_POOL = [
-    (3, 6, -2, 150.0, 19, 5, 2, -60.0),
-    (2, 5, 2, 150.0, 20, 6, -3, -60.0),
-    (5, 4, -3, 150.0, 17, 7, 1, -60.0),
-    (1, 7, 1, 150.0, 21, 4, -2, -60.0),
-    (6, 4, 1, 150.0, 14, 6, 3, -60.0),
-    (0, 5, 3, 150.0, 24, 4, -1, -60.0),
-    (4, 4, -1, 150.0, 12, 7, 2, -60.0),
-    (2, 6, -3, 150.0, 22, 4, 1, -60.0),
+    # Scanlines (see _glitch_scanlines below) repeat every 4px, so each
+    # band's height needs to be at least ~9-13px to reliably show 2-3 of
+    # them, regardless of the band's exact vertical alignment against the
+    # scanline grid -- much shorter and a band can land entirely between
+    # two scanline rows and show none at all.
+    (3, 12, -2, 150.0, 19, 11, 2, -60.0),
+    (2, 11, 2, 150.0, 20, 12, -3, -60.0),
+    (5, 10, -3, 150.0, 17, 13, 1, -60.0),
+    (1, 13, 1, 150.0, 21, 10, -2, -60.0),
+    (6, 10, 1, 150.0, 14, 12, 3, -60.0),
+    (0, 11, 3, 150.0, 20, 10, -1, -60.0),
+    (4, 10, -1, 150.0, 12, 13, 2, -60.0),
+    (2, 12, -3, 150.0, 22, 10, 1, -60.0),
 ]
 
 # render.Sequence doesn't loop on its own (see _glitch_scanlines below), so