Visit hoch.github.io/canopy and paste this code into the code window:

/ @channels 2
// @sampleRate 32768
// @duration 1

let nodeName = 'PannerNode';
let nodeOptions = {panningModel: 'HRTF', distanceMode: 'linear'};

// Merge channels for the destination.
let merger = new ChannelMergerNode(context, {numberOfInputs: 2});
merger.connect(context.destination);

let src = new ConstantSourceNode(context, {offset: 1});

// Impulse for testing.  We want a full buffer so as not to worry about
// the source disconnecting prematurely from the filter.
let b = new AudioBuffer(
    {length: context.length, sampleRate: context.sampleRate});
b.getChannelData(0)[0] = 1;
let impulse = new AudioBufferSourceNode(context, {buffer: b});

let testNode = new window[nodeName](context, nodeOptions);
let refNode = new window[nodeName](context, nodeOptions);

src.connect(testNode).connect(merger, 0, 0);
impulse.connect(refNode).connect(merger, 0, 1);

src.start();
src.stop(1 / context.sampleRate);
impulse.start();

Run it and examine the output waveforms.  In 65, channel 0 is silent. In ToT, channel 0 is not.

This is unexpected.