Should batchId start from 0?

1. A concise explanation of the problem you're experiencing.
I am working on my open source project to convert gltf to 3dtiles And here's my unfinished repository:
"Tileset.json" and the corresponding glb files(with batchIds) are successfully created. But I can not manage to do the correct picking in cesium.I
I looked into the code in cesium and found this snippet

My question is:
1. Does that indicate the batchId should start from 0?
2. How the picking works and how can cesium tell that vertices in different lods with same batchId belong to the same "feature"?

2. A minimal code example. If you’ve found a bug, this helps us reproduce and repair it.
var viewer = new Cesium.Viewer(‘cesiumContainer’, {
        geocoder: false,
        fullscreenButton: false,
        sceneModePicker: false,
        animation: false,
        selectionIndicator: false,
        baseLayerPicker: false,
        homeButton: false,
        infoBox: false,
        timeline: false,
        navigationHelpButton: false,
        navigationInstructionsInitiallyVisible: false,
        creditContainer: ‘dummy’,
        globe: false
    var tileset = new Cesium.Cesium3DTileset({
        url: ‘http://localhost:8081/tileset.json
    tileset.readyPromise.then(function(tileset) {
        viewer.zoomTo(tileset, new Cesium.HeadingPitchRange(0.5, -0.2,
     tileset.boundingSphere.radius * 2.0));
    }).otherwise(function(error) {


3. Context. Why do you need to do this? We might know a better way to accomplish your goal.

4. The Cesium version you're using, your operating system and browser.
1.52. chrome latest.

My test results show that it should be start at 0.

For your picking question, from what I can tell it looks like each type of content has a special pick shader or a texture it renders for the picking pass, this would be a good place to look:

Thank you for you reply. I've looked into the picking code. In my understanding, the batchId should be started from 0 and increment continuesly since cesium uses "batchId" to compute the uv to get the corresponding color of "pickTexture". I'm not sure if I'm right, but the experimental result seems good so far.