Vaadin Grid unusable on large datasets? (15 mill ++)


We are using Vaadin 8 with Grid. Datasets are lazy loaded using CallBackDataProvider.
Strange problem.
If count query returns 100000, then no problem scrolling.
If count query returns 500000, then things start to lag when scrolling, but its manageable.
But if count query returns 1000000 or 15000000 or more, then loading the grid takes forever (30 secs?) even though the fetch data query runs in a few milliseconds. And also, now the grid is basicly unscrollable. It lags and is unusable.

Now whats the problem here? With lazy loading there should not be issues regarding how large the dataset actually is?
Does Vaadin actually create xx empty rows from the count query and fills them with data lazily while you scroll?
For us, scrolling is not important. At least not beoynd the first 10000 rows. So can we make vaadin not create empty rows beoynd the first 10000?

This is not related to lazy loading. There is maximum number of rows (depending on browser and height of the row), before scrolling mechanism breaks, there is more discussion here about it:

Quoting: “For the record, the limit in Chrome seems to be 883 010 rows with the default rows height, even though it’s not possible to scroll back any more after reaching that far.”

Yes i see. We hardcoded the size to be maximum 40.000 if it was not less. It worked as a workaround