What

• Add unit tests for _detect_desktop() and for the mutual exclusion between IS_ANDROID, IS_IOS and IS_DESKTOP.

Why

• tests/test_utils.py covered _detect_ios() thoroughly but nothing exercised the desktop backend or the precedence logic at utils.py:105, so a regression there would have gone unnoticed. That logic is what keeps a component from falling through every platform branch during pn preview (the confusion described in FAQ on platform-specific code omits the browser preview (IS_WEB) #59).

How (brief)

• TestDesktopDetection follows the existing _detect_ios monkeypatch pattern: PN_PLATFORM=desktop returns True; unset, empty, another value, or a bare host platform returns False.
• TestPlatformFlagPrecedence covers _is_desktop = (not _is_android) and (not _is_ios) and _detect_desktop(). A _redetect() helper clears the cached _is_* globals and swaps in forced detectors, since all three flags are computed once at import time and there is no public seam to re-run detection through. monkeypatch restores both the caches and the real detectors at teardown. A parametrized sweep over all eight signal combinations asserts at most one flag is ever True.
• TestPlatformFlagsConsistency widened from two flags to three, including the PN_PLATFORM skip guard so the suite still passes when run under pn preview.

Testing

• pytest -q tests/test_utils.py passes (27 tests, up from 9); ./scripts/check.sh passes.

Risks/Impact

• Tests only, no source changes, no behavior change.

Docs/Follow-ups

• None. The desktop documentation gaps are tracked separately in Platform API docs omit the "desktop" OS value and is_desktop #39 and FAQ on platform-specific code omits the browser preview (IS_WEB) #59.

Closes #40