Would salting both of these messages lead to md5 hashes that no longer match?
Even with salting MD5 is still far too efficient to compute to be strong for password hashing. It could be combined in an iteration framework which made it secure, but there are plenty of other hash functions (with better reputations) that would be a better choice.
On the other hand, using the same salt twice will change the hash if it's put before the data, but not if it's put after the data.
Given an MD5 collision, you can apply a common suffix and still have a hash collision. Applying a common prefix changes the state the hash process is in when it reaches the colliding data and this breaks the collision.