In-vessel reflooding during a severe accident in a PWR was analyzed in 2000–2002 by a group of experts with the aim to prepare a global IRSN R&D strategy to answer the corresponding pending safety issues. Indeed, water is today systematically injected if available during a severe accident in a PWR. However, knowledge on consequences of such an injection is not complete and answers are necessary for accident management in present PWR as well for design and safety analysis of future PWR: is in-vessel corium retention possible? What is the kinetics of hydrogen production? What is the reactor cooling system (RCS) re-pressurization? Is there a risk of steam explosion (steam explosion is not discussed in this paper)? What is the impact on source term? And more generally, how to optimize water injection? (When? How?) R&D needs of investigation of these aspects were identified. This should cover separate-effect and integral tests, as well as modeling and code development. The approach consisted first in updating the synthesis of knowledge, based on the multiple reports released in an international frame (OECD, European Commission (EC) Framework Programs (FwP), …), and then in focusing on a detailed re-analysis of the most important experiments CORA, QUENCH, LOFT-LP-FP2 and of TMI-2 accident, the latter two being directly related to debris coolability phenomena. Several out-of-pile experiments on debris bed coolability were also analyzed. A qualitative analysis of different possible core degradation scenarios was performed for French PWR, depending on operator actions or procedures. Simplified and mechanistic models were used to evaluate orders of magnitude of the phenomena in reactor conditions. A Phenomena Identification Ranking Table (PIRT) ranked the elementary phenomena with respect, on one hand, to their importance from the point of view of safety consequences and, on the other hand, to their level of understanding (often based on experts opinion). In particular, coolability of debris either in the core or in the lower plenum was identified as an important issue to be solved: uncertainties were underlined on debris characterization (size, distribution, composition, …) and on multi-D thermal–hydraulics in a debris bed. This ranking is fully consistent with the outcomes of the EURSAFE 5th FwP project. Existing or future experiments that could satisfy the needs were then identified, including for debris coolability POMECO, DEBRIS, STYX, etc. This specific issue will be analyzed in the frame of the Network of Excellence on severe accidents SARNET which has started in 2004 in the 6th FwP. Further model developments are being performed in the IRSN codes ICARE/CATHARE (detailed modeling) and ASTEC (simplified fast-running modeling), the latter being jointly developed with GRS.