Hey @giddie! What guarantees does this project provide regarding message order and visibility? That is to say, one of the common “gotchas” with Postgres is assuming that things like sequences can be relied upon to act like monotonic cursors. For example:
Process A | Process B
BEGIN | BEGIN
insert next_val('my_seq') | insert next_val('my_seq')
1 | 2
# do something that takes time | COMMIT
COMMIT
In this scenario, the value 2 will be visible outside the transaction before 1, and so if a consumer treats “oh I have seen message with id 2, therefore I am caught up to 2” they will miss messages.
How does this project avoid that issue?
