The timing of optimal policy to combat climate change is controversial: while some advocate a ``gradual slope'' in policy, others emphasize the importance of redirecting investments rapidly if we hope to meet a ``2 degree'' goal. We consider this question in the light of irreversible ``dirty'' and ``clean'' investments, such as in coal-fired and solar electricity generation. This leads to a ``irreversibility effect'': the knowledge of an increasing carbon tax will reduce investments in assets that pollute, and so reduce emissions in the short term. This contrasts with the well-known ``Green Paradox'' effects of such policy on the suppliers of fossil fuels, such as coal mines. So stranded assets play opposing roles, depending on whether these assets supply or demand fossil fuels. We also show that it is optimal to begin deployment of clean technologies early, if their cost decreases with their cumulative deployment (as has been clearly observed for solar power). Moreover, if the rate of deployment increases for other reasons, their optimal subsidy also increases: our ``acceleration effect''. Additionally, we show that the choice of the instrument in a second-best setting depends on the stringency of climate policy: policy - subsidies for innovation. We derive these results theoretically, and quantify optimal policies in a dynamic general equilibrium climate-economy model.