Here's one way to quickly detect thin defects in a spool of filament.

I have a couple bad spools of filament with both fat and thin defects. A skinny spot kills a print just like fat ones do. Fat ones can be found by drilling a hole in something and feeding the filament through, but it's not so easy and would take time to check an entire spool for thin sections with a caliper.

By happy accident I found this quick and dirty method and it worked pretty well.

My Afinia has a fixed bearing-to-drivegear distance and it won't feed filament thinner than 0.055 inches (1.5 mm).

First set the caliper to your minimum acceptable thickness. Tighten the screw so it can't change.
Clamp the caliper horizontally to some rigid vertical surface as shown.
Thread the filament end through the wide crotch (?) of the caliper jaws.
Then run it all through from the full reel to an empty reel.

If there's a thin spot (thinner than the set width of the jaws), the filament will either catch in the jaws of the caliper, or fall all the way down and out if it's really thin.
I marked bad sections with a piece of masking tape.
Raise the filament back up into the jaws or just free it from being stuck, reel in a foot or two, and continue.

I used a hand cranked spool-to-spool setup and it doesn't take long to get a feel for it. Guide the filament with one hand and crank with the other.
(Hand crank files I printed a year ago are included because at least one file upload is required to publish. I do not recall any details other than they printed OK)

[Edit: Slept on it and recall the cone-nut has zero clearance. Scale it up on X-Y about 2 or 3%. Took a torch to the handle end and bent it to where I wanted it. Tapped the end of the axle with a 8x1 tap to fit a short bolt, securing the handle.
Lay the "stands" flat to print. Select crank direction so the cone nut tightens rather than loosens. It needs a lock nut or something... I wouldn't design a new one like this... Buyer beware.]

I ran my two troublesome spools... took about an hour.. and no longer need it, but the principle might be developed into a real project if someone cared to.