Retrieving the predicted eye position at a given time?

ovrHmd_GetTrackingState() allows one to retrieve the predicted head position at a given timestamp (via HeadPose.ThePose). ovrHmd_GetEyePose() provides predicted eye positions at the predicted frame render time. I'm trying to work out the equivalent of ovrHmd_GetTrackingState() for eye positions.

Tracing back from ovrHmd_GetEyePose() through the LibOVR source leads to this function in CAPI_FrameTimeManager.cpp:

Posef FrameTimeManager::GetEyePredictionPose(ovrHmd hmd, ovrEyeType eye)
{
    double           eyeRenderTime = GetEyePredictionTime(eye);
    ovrTrackingState eyeState      = ovrHmd_GetTrackingState(hmd, eyeRenderTime);
        
    // Record view pose sampling time for Latency reporting.
    if (RenderIMUTimeSeconds == 0.0)
    {
        // TODO: Figure out why this are not as accurate as ovr_GetTimeInSeconds()
        //RenderIMUTimeSeconds = eyeState.RawSensorData.TimeInSeconds;
        RenderIMUTimeSeconds = ovr_GetTimeInSeconds();
    }

    return eyeState.HeadPose.ThePose;
}

Why does FrameTimeManager::GetEyePredictionPose() return a head pose instead of an eye pose? It looks like the HeadPose field of the struct returned by ovrHmd_GetTrackingState() is not a head pose at all, but either the left or the right eye pose depending on the supplied timestamp. But that... doesn't seem right to me, especially as FrameTimeManager::GetEyePredictionTime ignores the eye parameter if VSync is disabled:

double FrameTimeManager::GetEyePredictionTime(ovrEyeType eye)
{
    if (VsyncEnabled)
    {
        return FrameTiming.EyeRenderTimes[eye];
    }

    // No VSync: Best guess for the near future
    return ovr_GetTimeInSeconds() + ScreenSwitchingDelay + NoVSyncToScanoutDelay;
}

What am I missing?