I have concerns about following cases.

* f-string containing multiple equivalent subexpressions:

f'{x} {x}'

or

f'''
{x}
{x}
'''

* f-string containing escaped newlines:

f'''\
\
{x}\
\
'''

The tests look slightly verbose and fragile to me. If once the parser will produce more efficient tree (see issue26415), the tests may fail. Needed more tests (for cases mentioned above), and adding them will add more verbose code. Maybe just find the child (or several children) of specific type and check lineno and col_offset only for it and few its children? If this will make the tests simpler. But if tests can't be made simpler, we should live with this.