Relational quantum mechanics is still a niche approach, but I bet it's the way of the future. It is conceptually fertile. It allows to define reality as the locus of intersubjective agreement (cf. mid p.4), so that different realities may correspond in different observer constituencies, e.g. homeopaths and non-homeopaths, and dreams are the ultimate instance, as Schopenhauer did not quite state, of a one-man reality. In this setting reproducibility, the core value of the scientific method, is a tool to extend agreement on a phenomenon to a larger community. The reproducibility of phenomena and the corresponding reality may vary in strength across communities, from extremely robust, Hiroshima-type, to exceedingly weak, as the results supporting theories regarded as pseudo-scientific usually are . Realities, i.e. loci of intersubjective agreement, may also expand and contract in time, as the history of science teaches us through such vivid examples as the rediscovered water-featuring M'pemba effect. This framework allows to identify the conditions and the obstacles, such as unwitting or wilful use of different standards and semantic models, that may affect reproducibility and hence the scientific validity of a result. Different questions eliciting different answers, as well as semantic equivalence in the eye of the experimenter, are important factors in the practical implementation of reproducibility (see  for a vivid example). The core question is no longer "Is this result reproducible?", but "By whom is it reproducible?".
It is often loudly claimed that "extraordinary claims require extraordinary evidence". To me this is either meaningless crap or it boils down to building bias into the scientific method, where bias has no place. I suggest the following instead: "Extraordinary claims require REPRODUCIBLE evidence". Actually, all scientific claims require reproducible evidence. Or I won't believe them, that is. However, introducing double standards into the implementation of the scientific paradigm may be necessary to maintain the scientific status of "big science" projects, which relie on massive "a posteriori" data-filtering (see e.g. how they "modify things a bit" at CERN) and are so costly that they results can hardly be reproduced by uninvolved observers. The introduction of a separate status for "extraordinary" (i.e. non-mainstream) claims allows mainstream science to protect itself by introducing a separate set of rules. Such double standards allow the liquidation of outsiders' claims, without impacting mainstream research, safeguarding its considerable network of economic interests and, perhaps most importantly, its status of ideological flagship.