PostgreSQL Stored Procedures vs Functions – Key Differences

Officially, PostgreSQL only has "functions". Trigger functions are sometimes referred to as "trigger procedures", but that usage has no distinct meaning. Internally, functions are sometimes referred to as procedures, such as in the system catalog pg_proc. That's a holdover from PostQUEL. Any features that some people (possibly with experience in different database systems) might associate with procedures, such as their relevance to preventing SQL injections or the use of output parameters, also apply to functions as they exist in PostgreSQL.

Now, when people in the PostgreSQL community talk about "stored procedures" or "real stored procedures", however, they often mean a hypothetical feature of a function-like object that can start and stop transactions in its body, something that current functions cannot do. The use of the term "stored procedure" in this context appears to be by analogy to other database products. See this mailing list thread for a vague idea.

In practice, however, this distinction of function versus procedure in terms of their transaction-controlling capabilities is not universally accepted, and certainly many programmers without database bias will take a Pascal-like interpretation of a procedure as a function without return value. (The SQL standard appears to take a middle ground, in that a procedure by default has a different transaction behavior than a function, but this can be adjusted per object.) So in any case, and especially when looking at questions on Stack Exchange with a very mixed audience, you should avoid assuming too much and use clearer terms or define the properties that you expect.

The comment by a_horse_with_no_name is actually the answer, but since it was not posted as an answer, I'm not able to mark it so. I've completely forgotten about transaction isolation levels, I've asked the question in VoltDB forums and it appears that as long as make sure that stored procedures are local to nodes (sorry, this is a bit of a VoltDB terminology), that is, they don't process data across partitioned tables, this should work.