Until 2009, Swiss-Prot had an "exponential-like" growth but then it became clearly slower ( http://web.expasy.org/docs/relnotes/relstat.html ). I would assume that would be the result of saturating the annotation staff, would that be right?
As for TrEMBL, it had a similar behaviour (only in a bigger scale) until 2015, and then there was a huge drop ( http://www.ebi.ac.uk/uniprot/TrEMBLstats ). What would explain that?