The Wikipedia article on Relation Algebra notes that this is a formal system which has essentially the same expressive power as the three-variable fragment of first-order logic. Peano Arithmetic can be encoded in the three-variable fragment, so this means that the satisfiability of an arbitrary RA formula is undecidable. There is an unsourced but intriguing remark:
(N.B. The Boolean algebra fragment of RA is complete and decidable.)
What is meant by "the Boolean algebra fragment of RA"?
I have tried to track down references, but I don't have access to the Tarski and Givant book which is one of the main texts (and which seems relevant judging by Monk's 1989 review). Any hints, references, or pointers (especially to journal articles or lecture notes that are online) would be welcome.