As mentioned in the long discussion of feature request #100 (https://sourceforge.net/p/gretl/feature-requests/100/?page=1), we shall not forget about deciding whether to make the "exact" unbalanced RE method of Swamy/Arora the default for gretl. Currently, it is an option.
OK, on the devel mailing list Allin has given the reasons why it's not necessarily the best choice to make that the default. So gretl's default will stay the same as Stata's here. Therefore closing this.
As a reference, here is Allin's message to the mailing list: http://lists.wfu.edu/pipermail/gretl-devel/2017-October/008143.html