I would guess that something's tipped. If the two shafts are not parallel, then one side of the belt would have more tension on it than the other, and it would drift like that.
It's pretty easy for the worm block to get tipped - ie: not perpendicular to the the ring gear. That's part of the point of the shimming step in the doc. Better gear mesh, and as a byproduct of that, the worm block is more likely to be square with the motor, as well.
So... the simplest thing would be to revisit section 7.4 and take another run at squaring the worm block.
Also, in Section 7.3... is the middle adjustment screw on your worm block a setscrew, or a capscrew?
Mounts: ES PMC-8 EXOS2
Scopes: SV 102EDT, ZWO 60/280 Guide
Cameras: Nikon D5300, Altair GPCAM2 290M
Software: ASCOM, CdC, AstroTortilla, BYN Pro, Sharpcap, PHD2
Computer: Thinkpad x230, Win7Pro/64