We’re sending one of the problematic switches in for analysis. I’m done with mine and my conclusion is that the cages are out of tolerance, probably in at least one batch of switches. We have used Extreme original SFPs and had them stick in the cages in the X695 but working flawlessly in an X450-G2. If they find a field fix for it or if we need to RMA them all is anyone’s guess. Extreme seems to have accepted this as a defect in some respect.
If this happens to any reader out there, the trick to free the SFP is to gently push it upwards (if you have the SFP in front of you and it’s in the top row) and then pulling the SFP’s lever to release it. This displaces the SFP just a tiny little bit, enabling the “fork” in the SFP to get in under the retainer (metal “tongue” in the cage) and free the SFP. If there’s an SFP in the port under (or over) the stuck one, pushing up on the SFP is impossible. Instead, a very gentle twist with a screwdriver between the SFPs will lift the SFP that half millimeter that is needed.
There are at least three different mechanisms in SFPs (Extreme has them all in various models) that lock them in place and release them. Only one type is problematic in the defective cages, but this type is used in both some old 1000Base-LX/SX without SSH and in some brand new 10G optics.
Explanatory image of an SFP cage: