What is this? "Can't find record in '<table>'' on query."

[time e.less]

Got MySQL Cluster up and running. Turned on replication. Every hour we have a fairly large transaction that runs to update several rows (~160,000) of a table. After finally setting these parameters:

MaxNoOfConcurrentOperations= 180000
MaxNoOfLocalOperations= 200000

Things seemed to work fine for a while. But after a couple of hours, this error showed up in my "show slave status" output:

Last_Error: Error 'Can't find record in 'a'' on query. Default database: 'production'.
Query: 'update low_priority tab1 a set a.rank =
(select rank from tab2 b where b.id = a.userid)'

A search on mysql.com turned up nothing. A search on Google turned up something about ndb_select_rows. Is this a normal part of running MySQL Cluster?

A little info about what I've done:

I run 3 physical computers, two data nodes per computer, two SQL nodes, and a single management node. There are two node groups, spread out across the three computers. The total size of the database is ~6GB, and each computer has 6GB RAM, for a total of 18GB. Each data node gets about 2.7GB of space, for a total of about 16GB RAM. Two copies of the database are stored.

MySQL Cluster versions are 5.0.16 straight out of Debian 64-bit for Opteron with NDB support compiled in.