Sorry about the long video, I uploaded to drive to share
https://drive.google.com/file/d/1M-SWdCYkUSu0xcF1qmb6ZyvDUr6v1CwG/view?usp=sharing

Originally posted by @RRosio in #17000 (comment)

This confirms that the part where an error is thrown an rendering is interrupted is fixed, but suggests that for very long cells which use estimated heights the rendering range may be incorrect.

Originally posted by @krassowski in #17000 (comment)

I am unable to reproduce this reliably locally. It seems to be a Heisenbug.

I think I caught it once and then the debugging was hinting at [..] the scrollOffset being set incorrectly (thus the rendering range was not updated correctly). [...]

Ideally we would be able to write a galata test reproducing the problematic behaviour.

Originally posted by @krassowski in #17000 (comment)