In psychology this is referred to as the common ingroup identity model for reducing intergroup bias. There are indications that the effects achieved are only temporary in nature, and negative outcomes can result from the loss of the overridden identities.
It probably depends on the scale. If you try to form a new in-group identity over a short time in a smallish group against a background that reinforces the existing categories, then I can believe the effects will be temporary. However, if the group is larger, the background identities weakened and de-emphasized, and enough time given, I think it'd work.
My impression is that current "anti-discrimination" attitudes actually reinforce the existing categories, even more strongly and explicitly than had been done before. There's a lot more labeling and "reasoning" through labels going on in places where it seems very counterproductive.
