Tipping points imply the potential for populations to transition rapidly from one cultural tradition or norm to another. In applied domains, this suggests that delimited policy initiatives may incite sweeping beneficial changes in behaviour. The risk, however, is that ordinary processes like individual heterogeneity and identity concerns might restructure or even undermine tipping in ways policy makers do not understand. To examine this possibility, we implemented an experiment before and after the fractious 2020 election for U.S. President. Participants played coordination games in groups of either Republicans or Democrats. Repeated play allowed groups to develop local norms. Once a group had established a norm, we exposed a subset of group members to an intervention that promoted an alternative group-beneficial norm. In our control treatment, choice options were labelled with neutral symbols. In our identity treatment, options were labelled with partisan political images designed to activate group identities and associated affective responses. We test whether this simple relabelling generated extreme differences in cultural dynamics. We expected that with neutral labels, groups transition to new norms quickly after an intervention. Under political labels, however, we test whether it is possible for groups to persist in a state of chronic disagreement after intervention, with large opportunity costs being the net result. In short we test, whether tipping points are powerful but fragile. Do group identities accelerate beneficial changes in local culture when choice and identity are unrelated? Can even a trivial link between choice and identity destroy the effect entirely?