I have been experimenting with slowing down the decay of qubits, and as a control, I implemented a circuit with just identity gates for comparison. However, when I implement such circuits in Qiskit, no matter how many identity gates I put, I get about the same fidelity each time. For example, I implemented the following two circuits with 10 identity gates and 20 identity gates in Qiskit and ran it using ibmqx4:
And I get the the respective histograms of their fidelity:
Which is obviously very weird. However, when I run the same thing in the composer on the IBM Quantum experience, I get the following histograms instead:
As you can see, the results from the composer is more correct. Am I doing anything wrong, or is there a bug with the implementation of identity gates in Qiskit?