Good question. In February 2019 we were very close to release a beta1 (expected for March) but since then I've been too busy to devote much time to Eigen. Moreover, bitbucket's bad new regarding the deletion of all mercurial repositories next year will keep us a bit busy to figure out a safe migration plan.
IIRC, early 2019, the major blocker for 3.4 was the expected refactoring of the Quaternion class [1][2]. Without a clear consensus, we'll have to drop it for 3.4. So beside fixing a few potential regressions [3] we should be good to go.