Reviews are being done for peers and the (mostly unpaid) editors, not for the publisher. It's academic service for the scientific community.

Although it's the publisher who ultimately creates profit from the work.

However, would reviewers be paid, then the publisher would certainly charge the authors for this. No gain.

Then refuse to do it. People complain about the journals, then do free work for a for-profit.

Refuse the service to the community that one is embedded in?

I see how one can justify reviewing for certain publishers, e.g. because they publish predatory journals or their shady business practices, but refusing to review any papers would be pretty much freeloading on the scientific community. Sure that's not an option.

No, I mean refuse to review for any journal that isn’t open source?

You said not reviewing for predatory journals, so apply that to journals that Elsevier owns?