It's intended behavior, but it should be documented better.

@rkern Can I work on a PR to document this behaviour? In that case can you assign this issue to me?

What is "intended" in changing dtype of array, which values are not changed¿

Both arguments are converted to ndarrays through the usual method of np.asarray(). The default dtype for empty lists is float64. Once the arguments are converted to ndarrays, then their dtypes are compared to figure out the common dtype they can both be safely coerced to; int64 and float64 can both go to float64.

>>> np.asarray([]).dtype
dtype('float64')

If you want to work around this, even in the case of empty lists, be sure to coerce both of your arguments to ndarrays explicitly with dtype=int ahead of time.

>>> a = np.array([1, 2], dtype=int)
>>> b = np.array([], dtype=int)
>>> np.append(a, b)
array([1, 2])

For what it's worth, I would agree that for append casting to the first array would probably have been a nicer choice for NumPy, but that doesn't mean I think we should change it (at least not without a long warning which may be tedious).

There was discussion (maybe even a PR?) to add a dtype= argument, which would allow you to write np.append(arr, ..., dtype=arr.dtype).
np.concatenate already has that option and is really basically what np.append uses internally.
That could be reactivated.

(For the empty list case, the dtype= argument may still be awkward due to unsafe casting of the "inferred" float64.)

Improving the docs is also good of course.