Wait one second before hover prefetch

docs/architecture/design-decisions.md

@@ -158,7 +158,7 @@ v3 auto-queued a background `?n=all` fetch for every chart on group open; on a
 22-chart group that queued tens of megabytes nobody asked for, contending with
 the windowed fetches a user actually waits on. v4 removed the auto-warmup: full
 history loads only on a deliberate per-chart signal — a window-chip click, a
-~600ms hover dwell, or a pan/zoom into the unloaded region.
+~1000ms hover dwell, or a pan/zoom into the unloaded region.
 → `web/components/Chart.tsx`, `web/lib/chart-format.ts`.
 
 **Bounded windows render on the full-length virtual x-axis.**

docs/architecture/performance.md

@@ -182,7 +182,7 @@ per-chart signal:
   windowed → loading → complete → error/retry state machine, so the partial view
   is never silent and a failed load is retryable. Charts with fewer than 100
   commits are born complete and show no chip.
-- **A ~600 ms hover *dwell*** (`HOVER_DWELL_MS`) starts a silent prefetch at a
+- **A ~1000 ms hover *dwell*** (`HOVER_DWELL_MS`) starts a silent prefetch at a
   mid-tier priority (`HOVER_PREFETCH_PRIORITY = 500_000`, between idle background
   `0` and a direct `INTERACTION_FULL_PRIORITY = 1_000_000`), so a deliberate
   hover has data ready while a mouse sweep across the page fetches nothing.

web/components/Chart.loading.test.tsx

@@ -227,7 +227,7 @@ describe('Chart opt-in full-history loading', () => {
     vi.useFakeTimers();
     card.dispatchEvent(new Event('pointerenter'));
     await act(async () => {
-      await vi.advanceTimersByTimeAsync(599);
+      await vi.advanceTimersByTimeAsync(999);
     });
     expect(fetchCalls.some((u) => u.includes('n=all'))).toBe(false);
     await act(async () => {
@@ -246,7 +246,7 @@ describe('Chart opt-in full-history loading', () => {
     });
     card.dispatchEvent(new Event('pointerleave'));
     await act(async () => {
-      await vi.advanceTimersByTimeAsync(600);
+      await vi.advanceTimersByTimeAsync(1000);
     });
     expect(fetchCalls.some((u) => u.includes('n=all'))).toBe(false);
   });
@@ -275,7 +275,7 @@ describe('Chart opt-in full-history loading', () => {
     vi.useFakeTimers();
     card.dispatchEvent(new Event('pointerenter'));
     await act(async () => {
-      await vi.advanceTimersByTimeAsync(700);
+      await vi.advanceTimersByTimeAsync(1100);
     });
     expect(fetchCalls.filter((u) => u.includes('n=all')).length).toBe(before);
   });
@@ -315,7 +315,7 @@ describe('Chart opt-in full-history loading', () => {
     vi.useFakeTimers();
     card.dispatchEvent(new Event('pointerenter'));
     await act(async () => {
-      await vi.advanceTimersByTimeAsync(601);
+      await vi.advanceTimersByTimeAsync(1001);
     });
     vi.useRealTimers();
     // The `?n=all` upgrade drains through the full-history queue across several

web/lib/chart-format.test.ts

@@ -679,8 +679,8 @@ describe('clampRangeWindow', () => {
 });
 
 describe('hover-dwell prefetch constants', () => {
-  it('dwell is a deliberate ~600ms pause, not an accidental sweep', () => {
-    expect(HOVER_DWELL_MS).toBe(600);
+  it('dwell is a deliberate one-second pause, not an accidental sweep', () => {
+    expect(HOVER_DWELL_MS).toBe(1000);
   });
 
   it('hover-prefetch priority sits above background (0) and below direct interaction', () => {

web/lib/chart-format.ts

@@ -45,7 +45,7 @@ export const HOVER_PREFETCH_PRIORITY = 500_000;
 /** How long the pointer must rest on one chart card before the silent
  * full-history prefetch starts, so a mouse sweep across the page fetches
  * nothing while a deliberate hover has data ready by the time the user acts. */
-export const HOVER_DWELL_MS = 600;
+export const HOVER_DWELL_MS = 1000;
 /** Per-fetch timeout (ms) for the chart `?n=100` / `?n=all` requests. A stalled
  * request aborts at this bound instead of spinning the loading indicator
  * forever. 30s is generous headroom over a cold Vercel function first-hit