Almost as bad as building a table with multiple secondary indexes are use cases that require multi-key or unbound queries. Many developers just need to know exactly how many rows their 200 GB table contains, so they run an unbound query while selecting a count:
SELECT COUNT(*) FROM some_really_huge_table_that_will_timeout;
Cassandra is simply not designed to scan its entire contents and return them in a nicely-formatted report. Queries such as this on large tables will likely result in a timeout.