Presently, aria-rowindex states

Authors MAY also place aria-rowindex on all of the children or owned elements of each row.

but, it doesn't indicate why anyone would do this. The example provided to demonstrate this 'may' specifies the aria-rowindex on the parent row as well as its gridcell children, which doesn't seem particularly necessary or helpful?

So, is there a reason one would do this? It isn't clear from the spec. Even if each cell had the aria-rowindex on it instead of the parent row, I'd question why the effort was taken to apply that attribute on multiple cells rather on the single parent? Are the potentials for conflicts if the row has no specified index, but its children cells do which may not match the current indexing that's actually present in the DOM?

If this attribute does have a good use case for being used on cell / gridcell, then we should update the text to better call out what that is.

If it doesn't, then I'm not sure why this is allowed. But, also wonder if there isnt' something useful that could be done with it?

For instance, recently I was asked if setting aria-rowindex on a cell could mitigate the need for specifying rows all together. So someone could create a series of sibling cells, and then set the necessary rowindex on them to divide the cells into the necessary rows without having to split up the cells into different container elements.

That would be kinda useful, actually, and based on the current spec text, i can absolutely see why they'd think this was possible.