Recently, the so-called robust diagnosability of DESs against permanent loss of observations (RDPLO) has been introduced. In this regard, the language generated by the system is said to be robustly diagnosable if it is possible to detect the failure occurrence, within a bounded delay, even when some sensors permanently fail to communicate the occurrence of the events to the diagnoser. In this technical note, we extend the definition of RDPLO to the decentralized case, considering the disjunctive architecture, leading to the definition of robust disjunctive-codiagnosability against permanent loss of observations (RDCPLO). The technical note also addresses the issue of online implementation, and we propose an efficient scheme to carry out online robust decentralized diagnosis against permanent loss of observations. Other contributions of the technical note are the development of algorithms for the verification of the RDCPLO, and the computation of the delay bound for robust decentralized diagnosis.