Fix Fill Capacity exceeding DC slot limit due to double-counted failed racks

computeRacksFailed was incremented on production failure and never decremented
when repaired racks came back online, while repair cohorts also tracked the
same racks. This caused usedSlots to inflate past the DC capacity over time.

Fix: derive computeRacksFailed from repair cohorts each tick instead of
maintaining it as a running counter. Include repair cohorts in pipeline slot
accounting so all racks are counted exactly once. Also fixes power limit in
fillDCToCapacity to only count online racks (pipeline racks don't draw power).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

apps/web/src/components/dev/EventTriggersTab.tsx

@@ -71,11 +71,15 @@ function resetRackFailures() {
       ...cluster,
       campuses: cluster.campuses.map((campus) => ({
         ...campus,
-        dataCenters: campus.dataCenters.map((dc) => ({
+        dataCenters: campus.dataCenters.map((dc) => {
+          const repairCount = dc.deploymentCohorts.filter(c => c.stage === 'repair').reduce((sum, c) => sum + c.count, 0);
+          return {
             ...dc,
-          computeRacksOnline: dc.computeRacksOnline + dc.computeRacksFailed,
+            computeRacksOnline: dc.computeRacksOnline + repairCount,
             computeRacksFailed: 0,
-        })),
+            deploymentCohorts: dc.deploymentCohorts.filter(c => c.stage !== 'repair'),
+          };
+        }),
       })),
     }));
     return { infrastructure: { ...s.infrastructure, clusters: newClusters } };

apps/web/src/pages/InfrastructurePage.tsx

@@ -113,7 +113,7 @@ function DeploymentProgressBar({ dc }: { dc: DataCenter }) {
   const tierConfig = DC_TIER_CONFIGS[dc.tier];
   const maxCompute = maxComputeRacks(tierConfig.rackSlots);
   const pipelineRacks = dc.deploymentCohorts.filter(c => c.stage !== 'decommission').reduce((s, c) => s + c.count, 0);
-  const totalTarget = dc.computeRacksOnline + dc.computeRacksFailed + pipelineRacks;
+  const totalTarget = dc.computeRacksOnline + pipelineRacks;
   const pct = totalTarget > 0 ? (dc.computeRacksOnline / totalTarget) * 100 : 0;
 
   if (totalTarget === 0 && dc.status === 'operational') return null;
@@ -642,8 +642,8 @@ function DataCenterDetailView({ clusterId, campusId, datacenterId }: {
 
   const tierConfig = DC_TIER_CONFIGS[dc.tier];
   const maxCompute = maxComputeRacks(tierConfig.rackSlots);
-  const existingCompute = dc.computeRacksOnline + dc.computeRacksFailed
+  const pipelineCount = dc.deploymentCohorts.filter(c => c.stage !== 'decommission').reduce((s, c) => s + c.count, 0);
-    + dc.deploymentCohorts.filter(c => c.stage !== 'decommission').reduce((s, c) => s + c.count, 0);
+  const existingCompute = dc.computeRacksOnline + pipelineCount;
   const availableSlots = maxCompute - existingCompute;
   const sku = dc.rackSkuId ? RACK_SKU_CONFIGS[dc.rackSkuId] : null;
   const netSlots = networkSlotsRequired(existingCompute);
@@ -814,7 +814,7 @@ function DataCenterDetailView({ clusterId, campusId, datacenterId }: {
               ) : (
                 <>
                   <p className="text-sm text-surface-400">
-                    Retrofit swaps all {dc.computeRacksOnline + dc.computeRacksFailed} <span className="text-surface-200">{sku!.name}</span> racks to a new SKU.
+                    Retrofit swaps all {dc.computeRacksOnline + pipelineCount} <span className="text-surface-200">{sku!.name}</span> racks to a new SKU.
                     The DC goes offline during retrofit.
                   </p>
                   <div className="space-y-2">
@@ -906,7 +906,7 @@ function DataCenterDetailView({ clusterId, campusId, datacenterId }: {
       {confirmRetrofit && (
         <ConfirmModal
           title="Confirm Retrofit"
-          message={`This will decommission all ${dc.computeRacksOnline + dc.computeRacksFailed} ${sku?.name} racks and install ${RACK_SKU_CONFIGS[confirmRetrofit].name}. The DC will go offline during this process.`}
+          message={`This will decommission all ${dc.computeRacksOnline + pipelineCount} ${sku?.name} racks and install ${RACK_SKU_CONFIGS[confirmRetrofit].name}. The DC will go offline during this process.`}
           confirmLabel="Start Retrofit"
           onConfirm={() => { retrofitDC(datacenterId, confirmRetrofit); setConfirmRetrofit(null); }}
           onCancel={() => setConfirmRetrofit(null)}

apps/web/src/store/index.ts

@@ -377,8 +377,8 @@ export const useGameStore = create<Store>()(
 
         const tierConfig = DC_TIER_CONFIGS[dc.tier];
         const maxCompute = maxComputeRacks(tierConfig.rackSlots);
-        const existingCompute = dc.computeRacksOnline + dc.computeRacksFailed
+        const pipelineCount = dc.deploymentCohorts.filter(c => c.stage !== 'decommission').reduce((sum, c) => sum + c.count, 0);
-          + dc.deploymentCohorts.filter(c => c.stage !== 'decommission').reduce((sum, c) => sum + c.count, 0);
+        const existingCompute = dc.computeRacksOnline + pipelineCount;
         const available = maxCompute - existingCompute;
         const actualQty = Math.min(quantity, available);
         if (actualQty <= 0) return s;
@@ -424,14 +424,14 @@ export const useGameStore = create<Store>()(
         const dc = found.dc;
         const tierConfig = DC_TIER_CONFIGS[dc.tier];
         const maxCompute = maxComputeRacks(tierConfig.rackSlots);
-        const existingCompute = dc.computeRacksOnline + dc.computeRacksFailed
+        const pipelineCount = dc.deploymentCohorts.filter(c => c.stage !== 'decommission').reduce((sum, c) => sum + c.count, 0);
-          + dc.deploymentCohorts.filter(c => c.stage !== 'decommission').reduce((sum, c) => sum + c.count, 0);
+        const existingCompute = dc.computeRacksOnline + pipelineCount;
         const available = maxCompute - existingCompute;
         if (available <= 0) return;
 
         const sku = RACK_SKU_CONFIGS[skuId];
         const affordableQty = Math.floor(s.economy.money / sku.baseCost);
-        const powerLimit = Math.floor((tierConfig.powerBudgetKW - existingCompute * sku.powerDrawKW) / sku.powerDrawKW);
+        const powerLimit = Math.floor((tierConfig.powerBudgetKW - dc.computeRacksOnline * sku.powerDrawKW) / sku.powerDrawKW);
         const qty = Math.min(available, affordableQty, powerLimit);
         if (qty <= 0) return;
 
@@ -498,7 +498,8 @@ export const useGameStore = create<Store>()(
         if (eraOrder.indexOf(s.meta.currentEra) < eraOrder.indexOf(sku.era)) return s;
         if (sku.requiredResearch && !s.research.completedResearch.includes(sku.requiredResearch)) return s;
 
-        const totalRacksToRetrofit = dc.computeRacksOnline + dc.computeRacksFailed;
+        const pipelineCount = dc.deploymentCohorts.filter(c => c.stage !== 'decommission').reduce((sum, c) => sum + c.count, 0);
+        const totalRacksToRetrofit = dc.computeRacksOnline + pipelineCount;
         if (totalRacksToRetrofit <= 0) return s;
 
         const oldSku = RACK_SKU_CONFIGS[dc.rackSkuId];

packages/game-engine/src/systems/infrastructureSystem.ts

@@ -195,7 +195,6 @@ export function processInfrastructure(state: GameState): InfraTickResult {
         }
 
         let computeRacksOnline = dc.computeRacksOnline;
-        let computeRacksFailed = dc.computeRacksFailed;
         let dcRepairCosts = 0;
 
         // Process retrofit
@@ -349,7 +348,6 @@ export function processInfrastructure(state: GameState): InfraTickResult {
           const prodFailures = binomialSample(computeRacksOnline, effectiveRate);
           if (prodFailures > 0) {
             computeRacksOnline -= prodFailures;
-            computeRacksFailed += prodFailures;
             const repairCost = sku.baseCost * sku.repairCostFraction * prodFailures;
             dcRepairCosts += repairCost;
 
@@ -400,12 +398,15 @@ export function processInfrastructure(state: GameState): InfraTickResult {
         // Compute aggregates for this DC
         const location = LOCATION_CONFIGS[cluster.locationId];
         const tierConfig = DC_TIER_CONFIGS[dc.tier];
-        const totalRacksInDc = computeRacksOnline + computeRacksFailed;
-        const netSlots = networkSlotsRequired(computeRacksOnline);
         const pipelineRacks = updatedCohorts
-          .filter(c => c.stage !== 'decommission' && c.stage !== 'repair')
+          .filter(c => c.stage !== 'decommission')
           .reduce((sum, c) => sum + c.count, 0);
-        const usedSlots = totalRacksInDc + netSlots + pipelineRacks;
+        const computeRacksFailed = updatedCohorts
+          .filter(c => c.stage === 'repair')
+          .reduce((sum, c) => sum + c.count, 0);
+        const totalRacksInDc = computeRacksOnline + pipelineRacks;
+        const netSlots = networkSlotsRequired(computeRacksOnline + pipelineRacks);
+        const usedSlots = computeRacksOnline + pipelineRacks + netSlots;
 
         let usedPowerKW = 0;
         let dcFlops = 0;