You sure that’s what is happening, and it’s not just mounting a different snapshot/dataset being mounted “on top” ?

I’ve seen it happen, which is why I ask. Assume the root dataset is named pool0 and has set0 set1 and set1/set2 as child datasets.

Their mount points are as follows:

/pool0/set0

/pool0/set1

/pool0/set1/set2

Now, if somehow, say set2 gets unmounted.temporarily, and you save files to /pool0/set1/set2 while the data set is not mounted, it’ll actually put those files in the set1 dataset, under the set2 directory.

But, when you mount the pool0/set1/set2 dataset again, the files under the set1 dataset are hidden by the set2 child.

Am I explaining it well enough for you to follow along?

Make sure you don’t have some similar situation by temporarily unmounting any nested datasets and ls’ing their mount points.