assertion failed: self.is_char_boundary(n)

[Sarrus1]

This seems to happen due to a race condition:

When initially parsing files, we read those files from disk.
If the files are edited by the user before we are done parsing, we might try to insert characters that are out of bounds.

This might also be due to (re)loading the workspace containing an unsaved file.
If the disk content is different than the one in memory, we start with an incorrect text and subsequent modifications might fail.