Revert "ref(flags): register LD hook in setup instead of init, and don't check for initialization"

[cmanallen]

Mutating a class attribute on __init__ violates encapsulation and will lead to strange errors. We need to rethink how we want to implement this before we merge any code.

A simple reproduction of the issue:

>>> class X:
...     y = 0
...     def __init__(self, z):
...         self.__class__.y = z
... 
>>> a = X(1)
>>> b = X(2)
>>> X.y
2
>>> a.y
2
>>> b.y
2

Reverts #3890

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
13770	1	13769	4167

View the top 1 failed tests by shortest run time

tests.test_basics test_classmethod_tracing

Stack Traces | 0.148s run time

tests/test_basics.py:946: in test_classmethod_tracing
    assert fake_start_child.call_count == 1
E   AssertionError: assert 2 == 1
E    +  where 2 = <MagicMock name='mock.start_child' id='139907889185552'>.call_count

To view more test analytics, go to the Test Analytics Dashboard
📢 Thoughts on this report? Let us know!

[cmanallen]

@antonpirker Is this a known flake? ERROR tests/integrations/sanic/test_sanic.py - TypeError: 'type' object is not subscriptable

[aliu39]

ERROR tests/integrations/sanic/test_sanic.py - TypeError: 'type' object is not subscriptable

+1, There's a similar error at https://github.com/getsentry/sentry-python/actions/runs/12600825778/job/35120696927?pr=3888

[antonpirker]

Yes, Sanic is sometimes flaky

[antonpirker]

Oh, the new Sanic release from 2025-01-01 is breaking our tests, so not flaky. I will fix this.