There's a quota issue with running this full-text search against large blocks of text. I originally used the searchable model instead of db.Model for my Bloog app, but I removed it after getting quota errors when Google hosted. There wasn't a problem with local testing, just when using searchable models in the cloud.
It may be that each search term in a full-text index is held in a separate but hidden table and therefore large chunks of text create a large number of rows. This would explain why inserts would be slower and why quotas are more likely to be encountered.
