The last time I looked at this, years ago, it looked fun, I just never spent any time digging into it. With my recent playing around with making a barrel cam, I thought I'd actually look into it because I thought they looked cool.
This cam just oscillates 90deg back and forth to prove the modeling process, and since I modeled the depth (and width) exactly the size of the follower, you can see it’s perfectly tangent on the sides, and only interferes a tiny bit at the bottom. But I expected that, and to me it doesn’t matter anyway since you’d always put in a decent amount of clearance on the bottom anyway and besides the actual cutting tool would clean that up like the follower mimics. So, with normal clearance on the bottom, I think it's as close as you'd ever practically be able to model it without a true solid body cut. The sides are what's important, not the bottom, and they appear to be spot-on. It was a fun little mental exercise to get the blood/creative juices flowing. In the real file, you can drag the “CAM_PNT0” around dynamically and it will make the 8-follower part turn around the vertical axis of the cam and even turn around its own axis so the rollers that come in contact with the cam always remain tangent (no interference), but in this zip file of the 2 part STEP files, you can can't do that but you can create an assembly of them aligning the default CS's and it should position the follower exactly as show in the pic, allowing you to see how all 4 roller followers fit perfectly at the same time. Have fun!