> because we rewrote the history Don't rewrite shared history. As for how migrations are ran. Last place, each team/service had their own data store. Numbere

> because we rewrote the history Don't rewrite shared history. As for how migrations are ran. Last place, each team/service had their own data store. Numbered sql files and forward compatible changes only. Sometimes this meant application code would have to write to two locations and in a later update change the read location. Current gig, everything is managed by the django orm. Great when you have a single db; sucks to scale out to sharded tables, sharded db instances, and is a pain for migrating to new data stores.