Question about Vacuum and Replication failures in 9.3.5

I noticed that tables on my master PostgreSQL server were growing, and running vacuum full analyze on them actually made them even bigger.

At the same time, a slave PostgreSQL server had fallen behind in trying to replicate, and was stuck in constantly looping over ‘started streaming WAL from primary at…’ and ‘requested WAL segment …. has already been removed’. Once I stopped running the slave instance, I was able to manually vacuum the tables, and appears that auto vacuum is now able to vacuum as well. One table (for instance) dropped from 10Gb down to 330Mb after this operation. I don’t see anything about auto vacuum not able to acquire locks while the slave wasn’t able to replicate. I am unclear why a slave trying to continue streaming would block the auto vacuum, or is something else at play?

I did check, and no base backups were in progress at the time this occurred.

Thank you,
Joel Avni