* add stacktrace when recovering a panic * add logging to the migrations * use string format * add context around why we return stacktrace