Should all bonds be constrained in lipid bilayer simulations with CHARMM36?

GROMACS version: 2023
GROMACS modification: Yes

According to the documentation, only h-bonds should be constrained with CHARMM36. Does that apply to simulations of lipid bilayers, or should all bonds be constrained?

Constraining bonds to H atoms is appropriate for any simulation with the CHARMM force field, irrespective of the contents of the system. Do not constrain all bonds with CHARMM.