TST: time/tests/test_precision.py failed in pyinstaller (computed error is different depending on the order of the arguments)

First failing log (2022-03-13): https://github.com/astropy/astropy/runs/5525474634

Last successful log (2022-03-12): https://github.com/astropy/astropy/runs/5519547613

Looks like this test was added in #10373 . Any idea how to fix, @Zac-HD or @mhvk ? 🙏

https://github.com/astropy/astropy/blob/c7b0e928e82dc7a4e099124d5223700e5bb4cfe2/astropy/time/tests/test_precision.py#L313-L315

____________________________ test_two_sum_symmetric ____________________________

    @given(floats(), floats())
>   def test_two_sum_symmetric(f1, f2):

astropy_tests/time/tests/test_precision.py:314: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

f1 = -3.089785075544792e+307, f2 = 1.7976931348623157e+308

    @given(floats(), floats())
    def test_two_sum_symmetric(f1, f2):
>       np.testing.assert_equal(two_sum(f1, f2), two_sum(f2, f1))
E       AssertionError: 
E       Items are not equal:
E       item=1
E       
E        ACTUAL: nan
E        DESIRED: -9.9792015476736e+291

astropy_tests/time/tests/test_precision.py:315: AssertionError
----------------------------- Captured stdout call -----------------------------
Falsifying example: test_two_sum_symmetric(
    f1=-3.089785075544792e+307, f2=1.7976931348623157e+308,
)

[Zac-HD]

I think there are two reasonable ways to fix this, though I don’t know which to prefer here:

1. check for and handle overflow inside the code under test
2. pass floats(min_value=1e100, max_value=1e100) (for example), so that we don’t try to test with overflow.

[pllim]

Thanks for the clarification, @Zac-HD ! I re-opened the issue and marked it as a real bug.