Why does 74hc595 work "better" without GRND (pin 8) grounded?

Here's the 7seg redone on one board with all wires in place.
I'm sure we can keep improving it, but I'm still at a loss for what could possibly be prohibiting at least some expected behavior.
Current behavior is that it goes on for a second, then off forever.



Here's the current code:

byte data = 11;
byte clk = 13;
byte latch = 10;
byte status = 13;

byte max_noop = 0x00;
byte max_d0 = 0x01;
byte max_d1 = 0x02;
byte max_d2 = 0x03;
byte max_d3 = 0x04;
byte max_d4 = 0x05;
byte max_d5 = 0x06;
byte max_d6 = 0x07;
byte max_d7 = 0x08;
byte max_decode_mode = 0x09;
byte max_intensity = 0x0A;
byte max_scan_limit = 0x0B;
byte max_shutdown = 0x0C;
byte max_test = 0x0F;

void output(byte address, byte data) {
  digitalWrite(clk, LOW);
  shiftOut(data, clk, MSBFIRST, address);
  shiftOut(data, clk, MSBFIRST, data);
  digitalWrite(latch, HIGH);
  digitalWrite(latch, LOW);
}

void setup() {
  Serial.begin(9600);

  pinMode(data, OUTPUT);
  pinMode(clk, OUTPUT);
  pinMode(latch, OUTPUT);
  pinMode(status, OUTPUT);

  digitalWrite(data, LOW);
  digitalWrite(clk, LOW);
  digitalWrite(latch, LOW);

  output(max_shutdown, 0x01);
  output(max_decode_mode, 0x00);
  output(max_scan_limit, 0x01);
}


void loop() {

  output(max_d0, HIGH);
  delay(1000);
  output(max_d0, LOW);
  delay(1000);
}