Whats the observed need behind merge replication in the dev environment? I understand the need for devs to have a local copy they can mess with, run tests against etc, but I'm lost on why a full Publisher-Subscriber model is needed to synchronize DB state in a dev/test environment, and it seems to be causing you more problems than it may solve given the schema is going to be malleable for a few months.
If merge replication is not a hard requirement for the dev environment, I would suggest you replace it with an alternate method of distributing changes to the local copies. If the devs are working with a full copy of the DB anyway, I see no reason not to write a script that backs up the master copy on the dev server, then pulls that file down and restores it locally. Then, changes to that schema would be accomplished with change scripts, which can be run and tested locally before being applied to the master DB, then distributed on-demand with another run of the backup/restore script.
It's a slightly more manual process and an older way to work with DBs, but it seems far more palatable to me than breaking and re-establishing replication regularly. It'll require some collaboration to make sure devs aren't trying to make a backup at the same time or making conflicting changes to local copies that will blow up on the master copy; your devs ideally should be talking to each other anyway about this kind of thing, and you might make the script smart enough to look for a recent backup before generating another.
One more thought, don't know how feasible it is given your progress to date; it's not impossible to switch from DB-First to Code-First. The conversion is basically a hybrid process of Database First and Code First; the DB is reverse-engineered as a one-time operation to generate a model similar to DB First, but instead of EDMX files, the model is written out to source code files, and changes to those model files or to mapping conventions on the context can then be aggregated and applied to the schema as migrations in typical Code First style. Assuming you prepare the live DB for migrations as well (and have the live DB in the same state as the master Dev DB prior to the model generation), this even removes the requirement of a SQL compare and update; you just apply the migrations to the live DB, same as you would to any Dev instance. The only potential gotcha is that some migrations can be written destructively, so you have to make sure what you're about to apply isn't going to clear out all the fields in a renamed column.
