I think the problem isn't permissive licenses in this case. It's not like AWS is running Matrix servers themselves or so.

The government deployments to far have been highly specified (and non-federating) systems requiring a lot of customizations, little of which is suitable for upstreaming and for which these "non-contributing" companies were specifically recruited for in a competitive bidding procedure (that EMS either lost or chose not to compete in). They usually also require on-prem deployments making the EMS cloud service unsuitable.

I think what we see here is a mix of market miss-match, i.e. EMS wants to be a public service, but their actually potentially paying customers want closed walled gardens for highly specific use-cases and a VC funded over-growth of EMS as a company and thus unrealistic ROI expectations and staff expenditures.

I think what will happen is that their VC funders will force them to spin-off a government service company to compete in such bids and slowly by slowly the focus will become primarily this as a service to the general public is not sufficiently profitable and it is also hard to compete with Slack and Discord there.

And in the end all the open-source enthusiasts will be left standing in the rain as the open Matrix federation is so over-engineered and has been developed with a "make it work somehow, fix it later" mindset, that without significant developer resources it will break under its own weight sooner or later.