The fission gas release (FGR) is a key point that must be accurately assessed for fuel rod design and licensing under both normal and off-normal conditions. FGR is important input data in terms of both the radioactive source term relative to the consequences of a nuclear accident on the environment, and the driving force regarding fuel damage. Correctly predicting its release under these conditions remains an important R&D goal to well understand the FGR impact during an accidental scenario. The present work deals with FGR mechanisms during a LOCA type transient. A new model to understand FGR is proposed; it is compared to experimental results and the possible phenomenon responsible of FGR is discussed. The samples involved (14 different tests) come from a section of PWR UO2 irradiated fuels. They were subjected to annealing tests on the MERARG facility (CEA Cadarache). They consist of a temperature ramp up to 1200°C at 0.2°C/s. The release occurs by bursts according to two main contributions (one between 600°C and 800°C and the second one after 1100°C). The largest FGR occurs at 1100°C and is due to a thermal activated process which is still thoroughly investigated. The release burst between 600 °C and 800 °C cannot be explained by the same type of mechanism. According to a completely different way, FGR could be induced by the creation of defects themselves created by alpha auto-irradiation (i.e. during the storage of the fuel sample). FGR kinetics might be modelled thanks to a two-step process: a phase of gas clusters nucleation and growth followed by their release. This new simplified two-parameter model for describing aggregation kinetics can be borrowed from the Finke-Watzky mechanism. It is a process of slow continuous nucleation linked to an autocatalytic surface growth. Both steps happen simultaneously so that gas clusters are consumed and released. When experimental data are fitted with the model, a very good agreement is obtained. Even though first results are promising, it is necessary to continue this study in order to confirm the preliminary conclusion.