This is a bit of a tracking issue: numpy is introducing __array_function__, with which one can override functionality of numpy functions for one's classes. To implement for:

• Quantity - finally, concatenate and friends will work! See Use __array_function__ to make most numpy functions work on Quantity #8808 and Quantity specific follow-up in Quantity.__array_function__ follow up for Quantity #8825.
• ShapedLikeNDArray - to cover shape-changing functions things like np.atleast_1d (Allow numpy shape functions on ShapedNDArray subclasses #10337)
• Extend ShapedLikeNDArray to np.where, np.concatenate and friends (like np.hstack, np.r_, etc.)
• For Time, make actual implementations for functions that currently work by creating time object array (e.g., np.sum on TimeDelta, which is used in gammay [TimeDelta does not support np.sum anymore #10501]; also np.diff on Time, and np.median [requested in Add Column.value as an alias for Column.data for consistency with Quantity and Time #10962]).
• Extend Time with other functions that logically can work, such as np.linspace (which would fix Make it possible to easily generate time series #8596) and np.isclose (relevant for Time equality comparisons are possibly too precise #8742)
• And use it in NDData, where probably many more functions make sense.