I expected my little error-correction test on a few qubits to hold up once I added basic repetition codes, but the actual runs showed logical errors climbing almost as fast as the physical ones the moment I tried to stretch beyond two gates.
The noise from the control lines and the readout crosstalk just wouldn’t stay low enough, no matter how I tuned the pulses.
Is it worth the effort and cost to chase better isolation hardware, or should I just live with the noisy results for now?