A bit of both. From a licensing point of view, Apache requires an Individual Contributor License Agreement (ICLA) from contributors, as well as other formalities, before contributions can be accepted. This is a non-starter given that LibreOffice (and go-OO before it) were in large part a reaction to the previous requirement for copyright assignment in the old OOo. So it wouldn't be possible for the LibreOffice code to be donated en masse to Apache, at least not without getting every contributor to *again* submit to relicensing their code (TDF is tri-licensed under the MPL/GPL/MGPL, which required obtaining a relicensing statement from everyone who had contributed code to OOo which hadn't been assigned to Sun / Oracle).
Then there's the addition problem of using AOO as a "clean start" once it's released, and rebasing all of the LibreOffice code on top of an Apache-licensed core. This appears to be getting held up primarily due to Apache bureaucracy: there are various CWSes (temporary feature branches) in the old OOo source which have unclear licensing (as only released source was explicitly relicensed), and the Apache mentors have been dismissive of requests to ensure that these are covered by the license grant because the requests don't come from signed-in AOOO "Committers".  LibreOffice may depend on source in those CWSes and cannot distribute them under its tri-license unless they are first established to be under the Apache license.
Actually *joining* AOO as contributors is seen as dangerous for the independent future of LibreOffice / TDF (a process which took over a decade from the first call for an independent OOo foundation) as there are parties (in particular one IBM staffer) who would inevitably try to ensure that as much press as possible was given to "TDF joining Apache" and thus popular perception being that TDF's "fork" was no longer important.
Politically, the general perception appears to be that Apache is operating as a convenient foil for IBM to usurp the OpenOffice brand name and community. The vast majority of actual developers active on the main AOO list at the moment appear to be Chinese IBM staff. IBM's promised code drop of their private IAccessible2 work still hasn't materialised (see  for a discussion last year: IBM has promised to release this code for five years) and nor has their Symphony user interface code (also promised for "some time after the AOO release", though reading between the lines of what IBM staff has suggested for "AOO 4.0" it would seem more likely that this would occur through IBM throwing an entire release over the wall once it's done, and not in a way that makes reusing that code for LibreOffice practical). All of this is anathema to the original goal of TDF, which was a community-driven OOo not stymied by overbearing bureaucracy or held to ransom by vendors with their own interests.