I have tried wrapping the calls with something like npm bottleneck, but it still seems that Prisma tries to batch them together