Pace back to 1.2x, and unweld the racket from over his head

PACE 1.3x -> 1.2x on all four numbers: 19.2 / 14.4 rally, 22.8 / 26.4 serve.

The strength-to-velocity relationship you suspected was already there is: a
splatter serve runs 26 m/s at the minimum pull, 36 at the kill threshold and
49.2 at a full draw. The draw very nearly DOUBLES the pace (1.89x), because
serveQualitySpeed carries more of the pair than serveSpeed does. Nothing to
change unless you want it steeper, and that is the one knob.

THE RACKET-ALWAYS-UP LOOK WAS A LATCH, not the art. hitServe sets pose 'serve'
with poseT 0; that only expires after POSE_TIME (0.38 s); and p_pull, the
wind-up frame, is only ever selected from pose 'idle'. So firing faster than
0.38 s reset the timer before it could expire, he never returned to idle, the
pull frame was never drawn once, and he stayed welded into p_smash — arm
extended overhead with balls appearing out of the strings. Exactly the late
game, because that is where the fast successions are.

serveToss now clears a stale 'serve' pose: a new wind-up ends the last shot's
follow-through. At ~4.6 shots/sec the cycle is now pull x5 frames, smash x8,
pull x5 — a real swing. Slow fire still plays the whole ritual: bounce, pull,
smash, idle, bounce.

Caveat: past roughly 8 shots/sec each pose gets 1-2 frames and will read as a
flicker rather than a swing. There is no fix for that short of drawn
follow-through frames; the poses are stills.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

config.js

@@ -10,13 +10,14 @@ export const TUNE = {
   reach: 1.3, aiReach: 1.6, // tight contact radius: the assist positions you, the tap must be honest
   swingWindup: 0.06,       // s before contact window opens
   swingWindow: 0.16,       // s contact window length
-  // ALL FOUR PACE NUMBERS ARE 1.3x THEIR ORIGINALS (16/12 and 19/22) — the whole ball, tennis and
+  // ALL FOUR PACE NUMBERS ARE 1.2x THEIR ORIGINALS (16/12 and 19/22) — the whole ball, tennis and
   // splatter, his and theirs, moved together so nothing about the relative feel shifts.
-  baseShotSpeed: 20.8, qualityShotSpeed: 15.6, // horizontal m/s = base + q*quality; perfect timing = real pace
+  baseShotSpeed: 19.2, qualityShotSpeed: 14.4, // horizontal m/s = base + q*quality; perfect timing = real pace
   // the slingshot's pace range. Weighted hard toward QUALITY so a full draw is a smash you can
-  // see: 27 m/s floppy at the minimum pull, 53 m/s at full. (baseShotSpeed above is the rally
-  // volley — shared with tennis mode and the CPU, and scaled with them.)
-  serveSpeed: 24.7, serveQualitySpeed: 28.6,
+  // see: 26 m/s floppy at the minimum pull, 49 m/s at full — the draw nearly DOUBLES the pace,
+  // which is the "harder pull, faster ball" relationship, and it is in serveQualitySpeed's share
+  // of the pair. (baseShotSpeed is the rally volley — shared with tennis and the CPU.)
+  serveSpeed: 22.8, serveQualitySpeed: 26.4,
   serveSpin: 1.3,   // extra gravity a full-power serve carries (was a fixed 0.35). Without it the
                     // net-clearance loop ate the whole smash: a flat fast ball can't get over the
                     // tape AND come down on a target, so it got throttled back to walking pace.