_Each_ INSERT/UPDATE/DELETE causes _all_ entries in the QC for that table(s) to be invalidated (purged). If you have done a lot of different SELECTs, and you have a gigabyte QC, then a lot of work is needed. (Apparently AWS Aurora has rewritten the QC code, so large query_cache_size works efficiently.)

My advice is to either turn off the QC (for production systems with lots of write traffic) or keep the size below 50M.

On a related topic, http://bugs.mysql.com/bug.php?id=21051 has improved the reset query cache operation and caused it to be less likely to block other queries in 5.0.25 and 5.1.12 and later. A reset might be less painful than a flush before a rare table update, even though it’ll remove all of the cached queries.

The query cache design will undoubtedly be improved to better suit large cache sizes at some time. For the moment, it’s intended for fairly small cache sizes, in tens rather than hundreds of megabytes.

The problems with extreme number of tables is another story.