In such extreme case (which, by the way, are disturbingly frequent at larger, very expensive commercial publishers) I recommend sending an e-mail to the editor, including the proofs to show that you are not talking about a misplaced comma, and simultaneously answering the production that you do not allow them to proceed. Beware not to sign or click anything that might make them able to pretend you allowed them to send the paper into production (as it would be much more difficult to get things right then). Ideally (i.e. if you are not forced through a web platform) you would e-mail the production team with copy to the Editor in Chief or handling editor for the record, or the other way round.
I don't recommend trying to fix manually an absurd amount of problems in Galley proofs: 1. Many libraries put a lot of money to pay for this work, you and your institution should not pay with your time; 2. you are bound to miss some of the bad stuff, half your corrections might be ignored, and new problems could be added after galley proofs (I would not believe that if I had not seen it myself -- my first paper actually).
I don't recommend sending a new version and suggesting they start back from that. It is pretty unpredictable what they could imagine doing, and if they use the same process again it will not be pretty.
Any diligent Editor in chief, when receiving your e-mail, should reach out to the publisher's management to have things sorted out. Without management involvement, the production team will probably do what they are probably asked to by default: treat things quick and dirty.
Last, I hope you put your preprint on the arXiv. If you did, you can at least have this as a version of record (and Elsevier allows you to update it to a postprint, including referees' remarks). If you did not already put the preprint on the arXiv, then Elsevier's policies are too complicated for me to understand and remember.