Is this really a breaking change? I can't imagine any code using Topology.from_openmm(openmm_topology, unique_molecule=None) currently works

The OpenMM Topology object currently supports bond orders and there is an issue to add formal charges. If these are present, unique molecules can be perceived from the Topology directly without the need to manually specify them. This was supposed to happen much sooner, and can be accelerated if needed.

I'd argue that we don't need to change the behavior, but we should try to get OpenMM to prioritize adding formal charges and bond orders (compatible with OpenFF) where possible.

@mattwthompson If it were a higher-value change I'd overlook the fact that it's at the benign end of the "breaking" scale. But it might still technically break someone's Topology subclass, or do something that we haven't anticipated. And since the gain is so marginal I'd rather have this wait until an API breaking release is scheduled.