Resistive Plate Chambers (RPCs) cover a surface of about 4000 m 2 , equivalent to 16 m 3 of gas volume both in ATLAS and CMS experiments at the Large Hadron Collider (LHC). The use of a relatively expensive Freon (R134a) - based gas mixture makes unavoidable their operation in closed-loop gas systems. It has been observed that the return gas of RPCs operated in background conditions similar to those foreseen at LHC contains a large amount of impurities, which are potentially dangerous for the long-term operation of these systems. During the past few years several RPCs have been operated in the intense radiation field of the CERN Gamma Irradiation Facility (GIF) in order to study the production of typical impurities, mainly fluoride ions, molecules of the Freon group and hydrocarbons. A systematic characterization of cleaning agents has also been performed. Moreover, the tests suggest an optimized configuration of filters, currently under long-term validation at the GIF set-up. The new filter configuration optimizes the filtering capacity for H 2 O, O 2 and RPC typical impurities. An important feature of the new configuration is the increase of the cycle duration for each purifier, that results in better system stability, reduced system downtime and, if needed, it permits to increase comfortably the gas flow in the detectors during the high luminosity running periods at LHC. The filtering optimization studies are complemented with a finite element simulation of the gas flow distribution in the RPCs, aiming at its eventual optimization. A preliminary study on the standard configuration for the RPC gas distribution shows regions in which the gas velocity is 10-100 times lower than in others. With a gas flow of 1 volume exchange every 4 hours (considered the lower limit for a safe operation without radiation) these regions represent about one third of the whole detector surface. The extension of these areas increases dramatically when decreasing the gas flow. A new RPC prototype with a flexible distribution of gas inlets and outlets has been built in order to experimentally quantify the impact of those critical regions on the detector performance and also to verify if a new, more effective solution can be found. The basic idea is that the regions with very low gas velocity define the overall gas flow needed for the detector operation. An efficient removal of the "used" gas mixture inside the RPC volume would permit the reduction of the overall gas flow rate and thus would cut down the operation cost of the large gas systems, without affecting the RPC detector's performance.