This term generally refers to allowing large areas of land to 'naturalise' assuming that they will eventually reach the same state they would have been in, had humans not modified them. Or, at least, they would exhibit much greater levels of biodiversity and ecological robustness than the controlled areas they replace. The notion has been popularised by George Monbiot's recent book Feral. It often includes deliberate introduction of significant wild species such as wolves, beavers and bears.
Behind this debate are two policy principles regarding the natural environment, sometimes referred to, respectively, as 'Land Sharing' and 'Land Sparing'. Land Sharing is the idea that we actually use most of the land for practical purposes (largely farming for food), but do it in such a way as to minimise impact on wildlife, for example by organic methods, minimising agricultural chemicals, maintaining hedgerows, re-timing certain operations etc. Production and nature are combined.
Land Sparing in in some ways the opposite, using as little land as possible for practical purposes in order to maximise the areas that can be released for habitat or indeed for re-wilding. Production and nature are separated.
The ideological tension arises because, on the whole, using less land implies intensification, producing more per hectare; and this is usually done by increasing inputs of resources such as energy, machinery, fertilisers and pesticides and by removing obstacles such as trees, hedgerows or competitive species. The paradox is that intensive farming usually employs environmentally-damaging processes, but in principle allows more land somewhere else to be used for pure habitat purposes. Both approaches can be deemed to have environmental benefits, but the literature shows no decisive advantage to either strategy in all circumstances.
There is increasing pressure on parts of the earth's land surface to produce more food, and it's going to get worse. It is not credible that this 'new farming' is going to be especially respectful of local biodiversity. Probably a better strategy is to set aside certain areas for genuine 'rewilding' and monitor them for compliance. This favours land sparing and intensification. By far the largest land sparing process is reduction of grazing livestock, followed by reduction of other livestock, followed by the use of arable/food crops fo energy production. Intensification can be accelerated by greater use of protected cropping, provided the energy required is virtually zero-carbon: which will be increasingly possible.
Full 're-wilding' cannot be expected to return any area to some earlier state, but could generate new balances, equally vigorous. As George Monbiot and others have demonstrated, the presence of top predators can help maintain the health of large ecosystems by controlling populations of herbivores that compromise the primary production of the whole system.