perf(ui): avoid polling already-ready folder-panel controls

UITests/DocumentationExportUITests.swift

@@ -103,12 +103,19 @@ final class DocumentationExportUITests: XCTestCase {
                     app.typeText(parent.path)
                     app.typeKey(.return, modifierFlags: [])
                     let panel = app.dialogs["open-panel"]
-                    XCTAssertTrue(panel.waitForExistence(timeout: 5), app.debugDescription)
+                    // Check current readiness before starting a polling wait; `||`
+                    // short-circuits, so delayed readiness keeps the same bounded waits.
+                    XCTAssertTrue(
+                        panel.exists || panel.waitForExistence(timeout: 5), app.debugDescription)
                     // AppKit exposes this button's localized AXTitle, not an AXLabel.
                     // Its in-dialog identifier excludes the Touch Bar duplicate.
                     let choose = panel.buttons["OKButton"]
-                    XCTAssertTrue(choose.waitForExistence(timeout: 5), app.debugDescription)
-                    XCTAssertTrue(choose.wait(for: \.isHittable, toEqual: true, timeout: 5))
+                    XCTAssertTrue(
+                        choose.exists || choose.waitForExistence(timeout: 5), app.debugDescription)
+                    XCTAssertTrue(
+                        choose.isHittable
+                            || choose.wait(for: \.isHittable, toEqual: true, timeout: 5),
+                        app.debugDescription)
                     recordSegment("panel")
                     choose.click()
                     XCTAssertTrue(

docs/DOCUMENTATION-UI-PROFILING.md

@@ -55,3 +55,22 @@ The performance lane's compile/link overhead is a separate hypothesis. Inspect t
 actual build graph and configuration differences before proposing reuse. Existing
 lane-partition checks already prevent duplicated performance/golden coverage work;
 that existing behavior is not a new optimization.
+
+## Already-ready folder-panel checks
+
+The folder-panel journey checks the current `exists`/`isHittable` state before
+starting XCTest's polling waits. The same three assertions and five-second
+fallback waits remain: the real open panel must exist, its scoped OK button must
+exist, and that button must be hittable. Export still uses the native button after
+Go to Folder keyboard navigation and path entry; no pasteboard shortcut, seeded
+panel location, or programmatic export replaces that interaction.
+
+Each check is a short-circuiting `exists || waitForExistence(...)` (or the
+`isHittable` equivalent), so an already-ready control skips the wait call and a
+delayed or missing control reaches the same bounded wait and failure. It is a
+candidate optimization, not a guaranteed duration reduction: XCTest's waits also
+evaluate readiness when they start, and the extra accessibility query adds cost
+when readiness is delayed. Use the identical eight-phase markers and matched-run
+protocol above to determine its effect. A single candidate CI run versus a
+historical baseline is diagnostic; it cannot establish a reproducible improvement
+or quantify workflow savings.