Display: Width of the image scroll slider is incorrect when samples_per_plugin is large

TensorBoard version: 1.13.0.
TensorFlow version: I am using PyTorch 1.0.0 with TensorBoardX, but since it is a display issue, the backend should be irrelevant.
OS: Linux, 5.0.5-1-MANJARO
Python version: 3.7
Browser: Chrome 73.0.3683.103

I am running an adversarial attack with 3000 iterations, and I want to visualize the current image in each iteration (since a rapid change may occur between two iterations). The default setting only shows images in 10 steps, which is too sparse for me.

When I run tensorboard --logdir attack_mnist/ --samples_per_plugin=images=3000, the slider is rendered as follows:

Note that 2999 is the maximum value.

For reference, if I run tensorboard --logdir attack_mnist/ --samples_per_plugin=images=300, the display is normal:

If I run tensorboard --logdir attack_mnist/ --samples_per_plugin=images=500:

If I run tensorboard --logdir attack_mnist/ --samples_per_plugin=images=1000:

It seems that the width of the scroll slider is getting larger, while the range of the slidable area keeps the same.

Reproduce the image display with samples_per_plugin=images=3000 in Chrome and compare it with 300, 500, and 1000. Trace the image plugin's slider rendering, then verify that the slider width and usable range remain correct for large sample counts.