fix(stars): focused-star sphere falls back to the focus slot

docs/BACKLOG.md

@@ -42,7 +42,7 @@ Items with a **→ details** link have a full write-up in [`backlog/`](backlog/)
 
 ## Rendering
 
-- [ ] **Field-star body braided to `select` slot** `ready` **important** — `#focus=star-<id>` arrives at an invisible star and deselect-at-close-range blinks the body out; sphere gates on `select` while URL fills `focus` and the sprite is distance-retired in-shader. → [details](backlog/2026-07-21-star-body-selection-braid.md)
+- [ ] **Multi-star sphere presence** `deferred` — the field-star sphere is one-at-a-time (nearest wins); a Gaia-resolved double a few AU apart would leave the companion sprite-retired with no body. → [details](backlog/2026-07-21-multi-star-sphere-presence.md)
 - [ ] **Saturn ring brightness** `ready` — the ring reads too dim next to the new limb-darkened disc; retune ring albedo/exposure (surfaced in the planet-atmospherics per-body visual pass).
 - [ ] **Photoreal-Earth follow-ups** `deferred` — drift traps + fidelity gaps from plans A–E (equirect-uv mirror, setMap kind table, shared proxy-sphere idiom). → [details](backlog/2026-07-19-photoreal-earth-followups.md)
 - [ ] **Titan atmosphere** `needs-design` — minimal params-row-over-flat-sphere vs full Venus-style cloud-as-surface + limb treatment (needs a texture through the fetch/build pipeline). → [details](backlog/2026-07-19-titan-atmosphere.md)
@@ -51,7 +51,6 @@ Items with a **→ details** link have a full write-up in [`backlog/`](backlog/)
 - [ ] **Perf-harness findings: measured hotspots** `needs-design` — large tier ≈ 3× medium (blows 60fps alone), small slower than medium (unexplained), solar-system 16.9 ms with vertex-bound hdr·NEAR0 at 60%. → [details](backlog/2026-07-21-perf-harness-findings.md)
 - [ ] **Lower-res offscreen star-aggregate pass** `ready` — try `STAR_AGGREGATE_DIVISOR` 2 → 4 (`renderTargets.ts`); ~4× further fill cut if the upsampled glow field survives visually.
 - [ ] **Bright star clump at ~5.9 kpc** `deferred` — flux verified conserved; residual over-exposure is display policy (mid-anchor slider + summed knee shipped; retune or tone-map shoulder next). → [details](backlog/2026-07-17-star-clump-brightness-5-9kpc.md)
-- [ ] **Field-star resolve predicate duplicated** `ready` — `starResolves` (focusedFieldStarSphereLayer) restates partitionStarsByResolution's per-star diameter+resolve rule; extract a shared `starResolvesToSphere` util both call.
 - [ ] **Foreground body draw/drawPick share a per-frame resolved set** `deferred` — mirrored partition/cull invocations can desync under future edits; star partition runs up to 4×/frame at deep zoom. → [details](backlog/2026-07-17-foreground-body-resolved-set.md)
 - [ ] **Star drawBudget small-tier mobile cap + iOS device pass** `deferred` — lower `hardCap` for `tier === 'small'` in `gaia-stars.ts`, tuned on a real device; verify the new vertex-stage storage bindings under WebKit's stricter WebGPU in the same pass.
 - [ ] **Star field → own slab** `needs-design` — the depthless star map (points/captions/connectors) inherits NEAR0's Earth-scale depth bracket; a STARS slab row deletes the three clip-z clamps + far-plane coupling. → [details](backlog/2026-07-13-star-field-own-slab.md)

docs/backlog/2026-07-21-multi-star-sphere-presence.md

@@ -0,0 +1,33 @@
+# Multi-star sphere presence (resolved double systems)
+
+**Surfaced:** 2026-07-21, designing presence-by-proximity (`fieldStarSphereLayer`).
+Deliberately shipped single-sphere; this is the extension if a real double ever
+matters.
+
+## The gap
+
+The layer draws ONE sphere — the nearest resolvable star (nearest-wins with
+hysteresis). A second catalogued star inside its own sprite-dissolve band
+(would-be sphere size 4–16 px, roughly 3–12 AU of camera distance for solar
+radius at 1080p) has its point sprite faded/retired but gets no sphere: the
+companion of a close pair goes invisible while the primary shows.
+
+Why this is acceptable today: the window only opens for two Gaia records within
+a few AU of each other. Gaia resolves pairs down to ~0.4–0.6 arcsec — a few AU
+of separation even at 10 pc — so such pairs are vanishingly rare in the bin,
+further thinned by magnitude subsampling. No known repro in the shipped data.
+
+## Extension sketch
+
+- `nearestResolvableStar` → `starsWithinRadius`: same expanded-box octree
+  descent, return ALL hits within the OFF radius (bounded small; the radius is
+  ~AU-scale).
+- Layer keeps per-star hysteresis over a small present-set (≤K, K≈4) instead of
+  one `PresentStar`; `draw`/`drawPick` loop the set — `starRenderer.draw` and
+  `drawFlooredSpherePick` are already per-sphere calls, so the loop is the only
+  change. Pick pass handles multiple spheres as-is (depth-tested nearest-wins).
+- All spheres share the one representative solar radius (no per-star radii in
+  the bin) — a true double renders as two identical suns.
+
+Design bits to settle: K and whether the ON takeover rule stays nearest-only or
+admits any star crossing ON (probably the latter once presence is a set).