simulation with bag play data get different results

edit

hi, I got a ros2 node, and the node subscribes topics (say /input) and publishes topics (say /output).

I also got a ros2 bag logged_input_in_run_mode contains the topics /input in real-time-run env, which i wish to make a simulation with bag play it as well as run with the node, to collect the /output in sim-mode(say the collected bag is logged_output_in_sim_mode), for offline analysis.

but i found the result (say some calculated velocity states) is different between logged_output_in_sim_mode and logged_output_in_run_mode, and even different between logged_output_in_sim_mode_1st_run and logged_output_in_sim_mode_2nd_run.

i.e. simulation with bag play data get different results...

i think it's something related to the time or clock issue when someone implements some part of the node, e.g. the node use some inner-clock-source instead of outer-clock-source (maybe for integral operation), and i cannot locate the code of this part, which is beyond my access.

so i tried clock and use_sim_time in the following shellscript to run the simulation and collect bag, but no improve:

ros2 run package node                                   &      
# the `node` is started in the same way and machine both in run-mode and sim-mode
ros2 param set /node use_sim_time True                  &
ros2 bag play --clock      ./logged_input_in_run_mode/  &
ros2 bag record /output -o ./logged_output_in_sim_mode  &

i understand the sequence of these cmd and time-delay between them may influence the results, but in my case it's insignificant and negligible.

my question is:

1. is my usage and understanding about --clock and use_sim_time right? for this kind of problem?
2. is there any other possible reason for this kind of problem? sim with bag data get different results?
3. is there any other method to make the simulated results consistent when play the same bag data and run the same version node?

thanks for any advice or discussion.