I just created a small labelmap volume, and populated it via the POST .../blocks endpoint. Here's the instance info:
http://emdata1:7600/api/node/2a128d8cc3ed4469acf4f609f0a25f14/gt-cube/info

The data I wrote extends from [64,64,64] to [576,576,576], but in the /info for the instance, DVID did not record the MinPoint correctly for all axes.

I tried deleting the instance and trying again, and this time the same thing happened, but with different incorrect values. In both cases, the max was set correctly.

Since this ingestion was performed in parallel from multiple workers, and since the erroneous MinPoint is not the same from one run to the next, it smells like a race condition problem.

Here's another example in the same repo:
http://emdata1:7600/api/node/2a128d8cc3ed4469acf4f609f0a25f14/segmentation/info

That data starts at [0,0,0], not [0,64,0]. (I can verify that the voxel data is stored correctly by viewing it in NeuTu or Neuroglancer.)