stim: why would order of CNOT operations affect detectors

I have two circuits (below) : first works fine (no complaints from stim); second comes back with a "non-deterministic detector" error. The difference between the two circuits is that the order of the first and second set of CNOTs is swapped; everything else is the same. I wasn't expecting the order of CNOT's to matter so this looks strange. Any explanation for it? circuit #1 : (works ok) R 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 H 27 28 29 30 31 32 33 34 35 TICK CX 0 18 1 19 2 20 3 21 4 22 5 23 6 24 7 25 8 26 27 0 28 1 29 2 30 3 31 4 32 5 33 6 34 7 35 8 TICK CX 3 18 4 19 5 20 6 21 7 22 8 23 0 24 1 25 2 26 27 2 28 0 29 1 30 5 31 3 32 4 33 8 34 6 35 7 TICK CX 10 18 11 19 9 20 13 21 14 22 12 23 16 24 17 25 15 26 27 15 28 16 29 17 30 9 31 10 32 11 33 12 34 13 35 14 TICK CX 9 18 10 19 11 20 12 21 13 22 14 23 15 24 16 25 17 26 27 9 28 10 29 11 30 12 31 13 32 14 33 15 34 16 35 17 TICK H 27 28 29 30 31 32 33 34 35 X_ERROR(0.01) 18 19 20 21 22 23 24 25 26 MR 18 19 20 21 22 23 24 25 26 DETECTOR rec[-9] DETECTOR rec[-8] DETECTOR rec[-7] DETECTOR rec[-6] DETECTOR rec[-5] DETECTOR rec[-4] DETECTOR rec[-3] DETECTOR rec[-2] DETECTOR rec[-1] TICK X_ERROR(0.01) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 MR 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 DETECTOR rec[-18] rec[-15] rec[-9] rec[-8] rec[-27] DETECTOR rec[-17] rec[-14] rec[-8] rec[-7] rec[-26] DETECTOR rec[-16] rec[-13] rec[-9] rec[-7] rec[-25] DETECTOR rec[-15] rec[-12] rec[-6] rec[-5] rec[-24] DETECTOR rec[-14] rec[-11] rec[-5] rec[-4] rec[-23] DETECTOR rec[-13] rec[-10] rec[-6] rec[-4] rec[-22] DETECTOR rec[-18] rec[-12] rec[-3] rec[-2] rec[-21] DETECTOR rec[-17] rec[-11] rec[-2] rec[-1] rec[-20] DETECTOR rec[-16] rec[-10] rec[-3] rec[-1] rec[-19] OBSERVABLE_INCLUDE(0) rec[-12] rec[-11] rec[-10] OBSERVABLE_INCLUDE(1) rec[-7] rec[-4] rec[-1] circuit #2 : (non-deterministic detector error) R 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 H 27 28 29 30 31 32 33 34 35 TICK CX 3 18 4 19 5 20 6 21 7 22 8 23 0 24 1 25 2 26 27 2 28 0 29 1 30 5 31 3 32 4 33 8 34 6 35 7 TICK CX 0 18 1 19 2 20 3 21 4 22 5 23 6 24 7 25 8 26 27 0 28 1 29 2 30 3 31 4 32 5 33 6 34 7 35 8 TICK CX 10 18 11 19 9 20 13 21 14 22 12 23 16 24 17 25 15 26 27 15 28 16 29 17 30 9 31 10 32 11 33 12 34 13 35 14 TICK CX 9 18 10 19 11 20 12 21 13 22 14 23 15 24 16 25 17 26 27 9 28 10 29 11 30 12 31 13 32 14 33 15 34 16 35 17 TICK H 27 28 29 30 31 32 33 34 35 X_ERROR(0.01) 18 19 20 21 22 23 24 25 26 MR 18 19 20 21 22 23 24 25 26 DETECTOR rec[-9] DETECTOR rec[-8] DETECTOR rec[-7] DETECTOR rec[-6] DETECTOR rec[-5] DETECTOR rec[-4] DETECTOR rec[-3] DETECTOR rec[-2] DETECTOR rec[-1] TICK X_ERROR(0.01) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 MR 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 DETECTOR rec[-18] rec[-15] rec[-9] rec[-8] rec[-27] DETECTOR rec[-17] rec[-14] rec[-8] rec[-7] rec[-26] DETECTOR rec[-16] rec[-13] rec[-9] rec[-7] rec[-25] DETECTOR rec[-15] rec[-12] rec[-6] rec[-5] rec[-24] DETECTOR rec[-14] rec[-11] rec[-5] rec[-4] rec[-23] DETECTOR rec[-13] rec[-10] rec[-6] rec[-4] rec[-22] DETECTOR rec[-18] rec[-12] rec[-3] rec[-2] rec[-21] DETECTOR rec[-17] rec[-11] rec[-2] rec[-1] rec[-20] DETECTOR rec[-16] rec[-10] rec[-3] rec[-1] rec[-19] OBSERVABLE_INCLUDE(0) rec[-12] rec[-11] rec[-10] OBSERVABLE_INCLUDE(1) rec[-7] rec[-4] rec[-1] error : ValueError: The circuit contains non-deterministic detectors. Circuit stack trace: during TICK layer #1 of 7 at instruction #1 [which is R 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35] There is no reason these groups of CNOTs should commute, because in both of them qubit $0$ is the control of a CNOT ($0\to24$ / $0\to18$) and the target of another one ($28\to0$ / $27\to0$). So the differing behaviour is not surprising. My advice is to never use a qubit more than once in a single CX command as it tends to reduce legibility. You can always merge them back at the end once the circuit works if you fear having too many commands might have a negative impact on the performances.