Identical servers different plan for the same query

[Eric Coll]

I have two identical servers running the same query but the query plans are completely different. How is this possible?

The servers are a master and a slave. The schema is identical compared by MySQL Workbench. The data is identical as it is replicated, and the server variables are identical as well.

The query is very simple:
select * from capacity.pool_usage pu
join capacity.last_timestamp lt
on pu.region = lt.region and
pu.timestamp = lt.timestamp and
lt.datatype = 'USAGE'
where
account in ('866211823536')

The master correctly grabs the timestamp record for each region in the lt table, and filters the usage table using a timestamp, regio index which is very efficient.

The slave queries all the usage for the selected account, and then joins with the lt table.

How is this possible? Shouldn't both servers use the same query plan?