stadium: wire STADIUM_CAPACITY_TICK in as a flat threshold, not a scheduler

Closes FABRIC-2.md's last open §12 Q5 question. fleet_heartbeat_tick_count
is fed by every live VM's own vm_tick(), not one VM's, so it was reaching
HEARTBEAT_INFERENCE_FREQUENCY (shared/borrowed from the per-VM inference
gate) several times faster than intended with more than one VM live -
backwards from FABRIC.md §22.4's required ~1000:1 separation.

What's actually gated turned out to be low-stakes: vm_physics_tick()
(capsule_vm_physics.c:397) is a passive statistics refit - re-sorts a
window of past heat-transfer samples and recomputes a median rate
estimate. It doesn't move heat or arbitrate capacity. Firing too often
just meant a noisier statistic recomputed more frequently than planned,
not incorrect behavior.

Considered and explicitly rejected: scaling the threshold by live VM
count at the check site. That's the first brick of a scheduler - reading
fleet state to adjust a rate dynamically - which this project has
deliberately avoided building. Implemented instead: STADIUM_CAPACITY_TICK
(existing Kconfig symbol, defined but never read by any code path) now
gates vm_physics_heartbeat_tick()'s call directly, replacing the borrowed
HEARTBEAT_INFERENCE_FREQUENCY. Default bumped 1000 -> 4000, a flat
constant picked once for Tripod's known 4-VM topology, same kind of
placeholder as every other frequency knob in Kconfig.kernel - not
computed from anything at runtime. Renamed fleet_last_inference_tick ->
fleet_last_capacity_tick to match. Still one clock, one counter
(fleet_heartbeat_tick_count) - just a bigger flat divisor on it.

Three-arch QEMU acceptance: all clean to ok>, identical Stadium
conservation invariant on all three (resident_sum=43691 reservoir=21845
sum=65536). logs/20260815-093425/amd64, logs/20260815-093521/aarch64,
logs/20260815-093641/riscv64.

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

FABRIC-2.md

@@ -763,34 +763,38 @@ all four `.claude/*.md` files, and correcting `.claude/CLAUDE.md`'s own pointer
   running both *after* Artemis lands, since Artemis's own storage/timing work is likely to
   perturb whatever gets measured now — a recommendation, not a ruling made here.
 
-**Needs Captain Bob's ruling:**
+- [x] **§12 Q5's `STADIUM_CAPACITY_TICK` wiring — CLOSED 2026-08-15.** A real, verified
+  ordering violation, live today: with Hermes restored (item 4.2) and `STADIUM_MAX_VM_COUNT`
+  (now RAM-derived, Section G) at more than one VM, the shared fleet counter reached
+  `HEARTBEAT_INFERENCE_FREQUENCY` several times faster than §22.4's required 1000:1
+  separation intended.
 
-- [ ] **§12 Q5's `STADIUM_CAPACITY_TICK` wiring.** A real, verified ordering violation, live
+  **What was actually at stake, checked before fixing anything:** the thing gated by this
-  today, not latent: with Hermes restored (item 4.2) and `STADIUM_MAX_VM_COUNT` defaulting to
+  threshold, `vm_physics_tick()` (`capsule_vm_physics.c:397`), is a passive statistics
-  4, the shared fleet-capacity counter can already reach `HEARTBEAT_INFERENCE_FREQUENCY` up to
+  refit — it re-sorts a sliding window of past heat-transfer samples and recomputes a median
-  ~4× sooner in wall-clock terms than a single VM's own heat-inference gate — backwards from
+  rate estimate. It does not move heat, does not arbitrate capacity, does not decide
-  §22.4's required 1000:1 separation, confirmed by this document's own §12 Q5 investigation
+  anything. Firing it too often just meant a noisier statistic recomputed more frequently
-  above.
+  than planned — not incorrect behavior, not a live bug in the urgent sense.
 
-  **Correction 2026-08-15 — the fix must stay inside the one-clock rule (`FABRIC.md` §16.4
+  **Fix, explicitly NOT the shape first proposed.** The initial framing ("threshold scaled
-  GAP-A1, §17.1 "two measures, one clock... this does not mean two clocks").** `fleet_
+  by live VM count") was rejected by Captain Bob before implementation — correctly: reading
-  heartbeat_tick_count` (`capsule_vm_physics.c:157`) is a single counter, incremented once
+  VM count at the check site to adjust a rate dynamically is the first brick of a scheduler,
-  per call to `vm_physics_heartbeat_tick()`, which itself only ever fires from inside the
+  which is exactly the kind of machinery this project has deliberately avoided building.
-  execution-paced `vm_tick()` (`vm_runtime.c:141`) — never from a hardware timer. So there is
+  Implemented instead: `STADIUM_CAPACITY_TICK` (existing Kconfig symbol, previously defined
-  only one clock here, exactly as ruled; the earlier framing of this item ("give the capacity
+  but never read by any code path) now gates `vm_physics_heartbeat_tick()`'s call to
-  tick its own named constant... independent of per-VM tick counts") was worded ambiguously
+  `vm_physics_tick()` directly, replacing the borrowed `HEARTBEAT_INFERENCE_FREQUENCY`.
-  enough to read as "add a second clock," which is not the fix and must not be built as one.
+  Default bumped from 1000 to a flat **4000** — a fixed constant picked once for Tripod's
-  The actual defect is narrower: `vm_tick()` calls `vm_physics_heartbeat_tick()` from *every*
+  known 4-VM topology, same kind of placeholder as every other frequency knob in
-  live VM's own dispatch (deliberate, `vm_runtime.c:126-138`'s own comment explains why — the
+  `Kconfig.kernel`, not computed from anything at runtime. Renamed
-  fleet-capacity signal needs to reflect aggregate activity, not just Hera's), so `fleet_
+  `fleet_last_inference_tick` → `fleet_last_capacity_tick` to match. Still one clock, one
-  heartbeat_tick_count` advances on the fleet's *combined* stream while `HEARTBEAT_INFERENCE_
+  counter (`fleet_heartbeat_tick_count`), unchanged — just a bigger flat divisor on it.
-  FREQUENCY` was written assuming a single VM's stream. Same one counter, same one clock —
+
-  the threshold compared against it just doesn't account for multiple VMs feeding it at once.
+  Files: `Kconfig.kernel`, `include/starforth_config.h`,
-  **Question for Bob:** the fix is to give `STADIUM_CAPACITY_TICK` a larger *threshold value*
+  `include/starkernel/vm/stadium.h` (comment only),
-  checked against the existing `fleet_heartbeat_tick_count` (still one clock, one counter —
+  `src/starkernel/capsule/capsule_vm_physics.c`. Three-arch QEMU acceptance, all clean to
-  just a bigger divisor, e.g. scaled by the live VM count so it restores something closer to
+  `ok>`, identical Stadium conservation invariant on all three
-  the required 1000:1 separation), not a second tick source. Wire it in that way, or resolve
+  (`resident_sum=43691 reservoir=21845 sum=65536`): `logs/20260815-093425/amd64/`,
-  the ordering some other way?
+  `logs/20260815-093521/aarch64/`, `logs/20260815-093641/riscv64/`.
 
 **Closed-by-ruling, not reopened this pass** (each already has a recorded "flag and leave"
 decision from Captain Bob — reversing that without being asked would be exactly the

Kconfig.kernel

@@ -64,17 +64,25 @@ config STADIUM_CONTAINS_DEPTH_MAX
 	  patrons held inside patrons within one Stadium.
 
 config STADIUM_CAPACITY_TICK
-	int "Capacity arbitration cadence, in virtual ticks (STADIUM_CAPACITY_TICK)"
+	int "Fleet transfer-slope re-estimation cadence, in virtual ticks (STADIUM_CAPACITY_TICK)"
-	default 1000
+	default 4000
 	help
-	  How often Hera re-evaluates capacity transfers between VMs
+	  How often vm_physics_heartbeat_tick() re-fits the fleet's
-	  (FABRIC.md item 1.4), expressed in virtual ticks -- never
+	  transfer-slope estimate (vm_physics_tick() -- a passive median
-	  wall-clock. Default matches the existing precedent at
+	  recompute over recent touch samples, not a capacity/transfer
-	  capsule_vm_physics.c's vm_physics_heartbeat_tick()
+	  decision itself), expressed in virtual ticks -- never wall-clock.
-	  (HEARTBEAT_INFERENCE_FREQUENCY, also 1000), comfortably past
+	  Wired in 2026-08-15 (FABRIC-2.md F.2/§12 Q5): this counter is fed
-	  the "order of magnitude apart from the heat tick" minimum
+	  by EVERY live VM's own vm_tick(), not one VM's, so it previously
-	  (FABRIC.md §22.4). No consumer yet -- capacity arbitration
+	  shared HEARTBEAT_INFERENCE_FREQUENCY (1000) and fired roughly
-	  itself is not yet on the punch list.
+	  (live VM count) times faster than a single VM's own heat-inference
+	  gate -- backwards from the "order of magnitude apart" minimum
+	  (FABRIC.md §22.4). Default of 4000 is a flat, untuned placeholder
+	  picked to roughly restore that separation at Tripod's known
+	  4-VM topology (Hera + two Hermes + Artemis) -- not computed from
+	  live VM count at runtime, deliberately: a fixed constant, same
+	  as every other frequency knob here, not adaptive logic. Real
+	  tuning is DoE work (item 5.1), same treatment as the other
+	  placeholder constants in this file.
 
 config STADIUM_MEMORY_PERCENT
 	int "Percent of free physical memory the Stadium claims at boot (STADIUM_MEMORY_PERCENT)"

capsules/BLOCK_MAP.md

@@ -1,5 +1,5 @@
 # Capsule Block Manifest — Auto-generated
-<!-- Generated by mkcapsule --manifest  2026-08-15T12:09:24Z -->
+<!-- Generated by mkcapsule --manifest  2026-08-15T13:36:13Z -->
 <!-- DO NOT EDIT — re-run mkcapsule --manifest to refresh.  -->
 <!-- Hand-written justifications and immutability notes live  -->
 <!-- in MANIFEST.md alongside this auto-generated index.      -->

include/starforth_config.h

@@ -72,7 +72,7 @@
 #define STARFORTH_CONFIG_HEARTBEAT_INFERENCE_FREQUENCY_DEFAULT 1000
 #define STARFORTH_CONFIG_STADIUM_VM_MEMORY_PERCENT_DEFAULT 50
 #define STARFORTH_CONFIG_STADIUM_CONTAINS_DEPTH_MAX_DEFAULT 5
-#define STARFORTH_CONFIG_STADIUM_CAPACITY_TICK_DEFAULT 1000
+#define STARFORTH_CONFIG_STADIUM_CAPACITY_TICK_DEFAULT 4000
 #define STARFORTH_CONFIG_STADIUM_MEMORY_PERCENT_DEFAULT 1
 #define STARFORTH_CONFIG_STADIUM_WORD_HEAT_QUANTUM_DEFAULT 2048
 #define STARFORTH_CONFIG_STADIUM_WORD_COOL_RATE_Q48_DEFAULT 21845

include/starkernel/vm/stadium.h

@@ -109,10 +109,12 @@ typedef char stadium_cell_size_check[(sizeof(StadiumCell) == STADIUM_CELL_BYTES)
 
 /*
  * Items 1.1 and 1.4 named this item as where their Kconfig symbols would be
- * implemented. Neither has a consumer yet (item 3.5 for the depth cap,
+ * implemented. STADIUM_CONTAINS_DEPTH_MAX still has no consumer (item 3.5
- * capacity arbitration -- not yet on the punch list -- for the tick); these
+ * for the depth cap, not yet implemented). STADIUM_CAPACITY_TICK was wired
- * checks only prove the symbols are defined and sane, the same discipline
+ * in 2026-08-15 (capsule_vm_physics.c's vm_physics_heartbeat_tick(), see
- * already applied to the byte-count checks above.
+ * FABRIC-2.md F.2/§12 Q5) -- this check now proves a real, live constant
+ * is sane, not just a placeholder, same discipline already applied to the
+ * byte-count checks above.
  */
 typedef char stadium_contains_depth_configured_check[(STADIUM_CONTAINS_DEPTH_MAX > 0) ? 1 : -1];
 typedef char stadium_capacity_tick_configured_check[(STADIUM_CAPACITY_TICK > 0) ? 1 : -1];

src/starkernel/capsule/capsule_vm_physics.c

@@ -39,7 +39,7 @@
 #include "starkernel/kmalloc.h"
 #include "starkernel/q48_16.h"
 #include "starkernel/console.h"
-#include "starforth_config.h" /* HEARTBEAT_INFERENCE_FREQUENCY */
+#include "starforth_config.h" /* STADIUM_CAPACITY_TICK */
 #include "starkernel/vm/bootstrap/sk_vm_bootstrap.h" /* sk_get_mama_vm */
 #include "vm.h"           /* VM, vm->ssm_l8_state */
 #include "ssm_jacquard.h" /* ssm_l8_state_t, CDConfigTable */
@@ -155,7 +155,10 @@ static uint64_t slope_fit_quality_q48    = 0;
  * without a forward declaration -- it is now this counter's second
  * reader, alongside vm_physics_heartbeat_tick()'s own gating use. */
 static uint64_t fleet_heartbeat_tick_count  = 0;
-static uint64_t fleet_last_inference_tick   = 0;
+static uint64_t fleet_last_capacity_tick    = 0; /* renamed 2026-08-15 from
+                          fleet_last_inference_tick -- gated on
+                          STADIUM_CAPACITY_TICK now, not HEARTBEAT_
+                          INFERENCE_FREQUENCY, see vm_physics_heartbeat_tick() */
 
 static vm_physics_node_t *vm_physics_find(VMUuid vm_id)
 {
@@ -454,12 +457,22 @@ void vm_physics_tick(uint64_t now_ns)
     fleet_transfer_slope_q48 = (rates[rate_count / 2] * l8_regime_modulation_q16()) >> 16;
 }
 
+/* Gated on STADIUM_CAPACITY_TICK, not HEARTBEAT_INFERENCE_FREQUENCY (fixed
+ * 2026-08-15, FABRIC-2.md §12 Q5): fleet_heartbeat_tick_count is fed by
+ * EVERY live VM's own vm_tick() (see the comment above this counter's
+ * declaration), so it advances several times faster in wall-clock terms
+ * than a single VM's own tick_count -- comparing it against
+ * HEARTBEAT_INFERENCE_FREQUENCY (written assuming a single VM's stream)
+ * meant this fired far more often than §22.4's required ~1000:1 separation
+ * from a VM's own heat-inference loop. Still one clock, one counter --
+ * this is a bigger flat threshold on the same counter, not a second tick
+ * source or anything that reads live VM count to adjust itself. */
 void vm_physics_heartbeat_tick(uint64_t now_ns)
 {
     fleet_heartbeat_tick_count++;
-    if ((fleet_heartbeat_tick_count - fleet_last_inference_tick) >= HEARTBEAT_INFERENCE_FREQUENCY) {
+    if ((fleet_heartbeat_tick_count - fleet_last_capacity_tick) >= STADIUM_CAPACITY_TICK) {
         vm_physics_tick(now_ns);
-        fleet_last_inference_tick = fleet_heartbeat_tick_count;
+        fleet_last_capacity_tick = fleet_heartbeat_tick_count;
     }
 }