When the crash happens, there are two sets of queries running on the server.
One set is populating Business Intelligence aggregate tables, the other is refreshing a development copy of a database, using parallel loading.
The two sets of queries are running on different, independent databases.
The crash does not occur when only one set of queries (e.g. only the db load queries, or only the BI aggregator queries ) are running.
The tables that I try to drop are in compressed format.
The system is IO bound when the crash occurs (iostat consistently shows 100% disk utilization), but not CPU bound.
The crash occurs when I try to drop several big (multi gigabyte) tables in parallel, as you can see in the attached error log.