LB particle coupling thermostat seed doesn't fully decorrelate independent simulations about espresso HOT 2 CLOSED

This topic was actually already brought up for all other thermostats in #3585. In fact, the solution implemented in #4845 is exactly the same as the solution we implemented back in 2020. I don't remember why we chose to leave the LB particle coupling code out of the bugfix PR back then.

from espresso.

Offline discussion with @RudolfWeeber: this is indeed a bug, albeit not a serious one, since it only affects ensemble runs.

I looked into the possibility of backporting the bugfix to 4.2, but the LB CPU and LB GPU use different Philox kernels while sharing the same RNG struct, thus significant alterations to the LB GPU code would be required. We will not backport it to 4.2.

from espresso.