Taking the following 2 code cells as an example:

Cell 1:

def foo():
	print("first")

Cell 2:

	print("second")

Currently, Ruff would concatenate it into a single source and pass it to the parser. This means that the parser would see it as:

def foo():
	print("first")
	print("second")

Which is a valid code. But, actually the indentation before the second print statement is invalid as it is in another cell.

A possible solution, as suggested by @MichaReiser, would be to run the parser per cell, merge the statements, and update the node ranges for all the subsequent cells with the correct offset.