[en] For 30 years, outputs published by UK researchers have been submitted for cyclical assessment by peer panels, creating a unique longitudinal dataset. This is the first analysis covering 921,254 submitted outputs and 36,244 case study references across 25 years, five assessment cycles and both academic impact and economic/societal impact. For submitted outputs, document types shift towards journal articles across time. The time-spread of outputs is skewed to the most recent publication years in early RAE cycles, a pattern not reported at the time but one that then changes synchronously for science and engineering but not for social science or humanities. The skew and later changes are cohesive across disciplines and institutions. For impact case study references, the time-spread of the earliest of the six references for each study is uniform for science and engineering, although the overall timespread for references is skewed. About 42% of case study references with DOIs can be identified as RAE/REF submitted outputs, at about the same rate in every publication year for 1996-2013. Implications for assessment are discussed. (Author)