Fractional unit root - solutions to?

[legitlicence]

There are loads of papers on tests yet no one has come up with a new standard approach for cointegration that makes it into textbooks, the history seems to have gone Engle-Granger, Johansen, proliferation of tests with everyone knowing unit root tests are not very reliable but carrying on anyway. I don't even know what good practice would be if I did one of the involved tests and found that the likeliest order of integration was for example 1.7. So question: what should a person do if if they saw a test was I(x) where x is between 0.5 and 1.9 but not 1 or really close to 1?

If the textbook writers can't find a serviceable response maybe eviews users and staff can.