increase queue by 16x
this should give a bunch of seconds of buffer for the writer to get through stuff before we let pressure reach back to the remote jetstream instance. *hopefully* this will lead to fewer reconnects due to getting booted by jetstream for being slow.