A decorator performing automatic method generation on a class by looping: all methods seem to use the same exact inputs but shouldn't

Question

Originally, this question was "deepcopy a string in python" and I am keeping the essential parts of the discussion here for reference. Although I did solve the issue, I do not understand neither the origin of the issue nor how the solution differs from the wrong/not working implementation. I do want to understand this and took my time trimming the actual code into a minimal working example.

But first, part of original content: Deep copies of strings return the exact same string:

import copy
_1= "str"
_2 = copy.deepcopy(_str)
_1 is _2 # True

Novel content: The questions are inline with the code. Basically, I am decorating a TestClass with a decorator that is going to generate standardized 'appender' methods on the class for every attribute inserted in TestClass.channels. Please note, I am not posting the question to discuss software architecture/implementation, but to actually understand the interstitials of Python relevant to the noted issue: why all 'appenders' generated will actually perform exactly the same work as that prepared for the last attribute of TestClass.channels defined:

import pandas as pd


def standardize_appenders(kls):
    # Checks `kls` for `channels`, whether an appropriate pseudo implementation of 'appenders' is
    # available and actually implements the `appenders`: BESIDES AUTOMATION, I AM FORCING A STANDARD
    channels = getattr(kls, 'channels')
    chs = [o for o in dir(channels) if not o.startswith('_')]
    chs_type = [getattr(channels, o) for o in chs]

    for _ch, _ch_type in zip(chs, chs_type):  # ISSUE SOMEWHERE HERE
        ch = _ch
        ch_type = _ch_type

        # THE `def appender(self, value, ch=ch, ch_type=ch_type)` SIGNATURE SOLVES THE PROB BUT I DON'T UNDERSTAND WHY
        def appender(self, value):
            # nonlocal ch, ch_type
            if not isinstance(value, ch_type):
                raise TypeError(f'`{ch}` only accepts values of `{ch_type}` type, but found {type(value)}')
            data: dict = getattr(self, 'data')  # let's not discuss the need for this knowledge please
            data_list = data.setdefault(ch, [])
            data_list.append(value)
            data.setdefault(ch, data_list)

        appender.__doc__ = f'<some {ch}/{ch_type}> specific docstring'

        setattr(kls, 'append_' + ch, appender)  # ALL APPENDERS OF `kls` WILL DO ESSENTIALLY THE WORK DEFINED LAST

    return kls


@standardize_appenders
class TestClass:
    class channels:
        dataframe = pd.DataFrame
        text = str

    def __init__(self):
        self.data = {}


if __name__ == '__main__':
    test_inst = TestClass()
    # test_inst.append_dataframe(pd.DataFrame({"col1": [1, 2, 3], "col2": list("abc")}))
    # TypeError: `text` only accepts values of `<class 'str'>` type, but found <class 'pandas.core.frame.DataFrame'>
    test_inst.append_dataframe("A minimal trimmed example")  # NOTE THE WRONG 'APPENDER'
    test_inst.append_text("of the original implementation")
    print(test_inst.data)
    # Out {'text': ['A minimal trimmed example', 'of the original implementation']}

Be my guest to propose a better title. I am out of good ideas for a short, but informative title for this case.

(Windows 10, python 3.8 from an anaconda env)