XDS used the same scan line area (“below” the VITS signals and “above” the program video) as the various captioning services, so not affected by VCR head switching. Teletext appeared “above” the VITS signals higher in the raster, where head switching takes place.
Are you quite certain? With the VHS and Betamax formats, it was always my understanding that the head switching occurred entirely within the usually-overscan-concealed final scanlines of the video portion of the frame, just before the vertical blanking interval began. That would have been the best "sacrificial zone" for the format designers to choose to locate any head switching signal disruptions. Not only because most viewers wouldn't be able to see that area, but because placing massive disturbances within the actual VBI would have trashed the -40 IRE synchronization pulses housed therein, making it difficult for televisions to maintain consistent vertical locks throughout playback.
I believe the actual reason for the difficulty people have recovering data in the VBIs of older recordings is mistracking. The VHS and Betamax formats' specs never dictated building machines to professional tolerances. Decks often could not play tapes recorded in other decks without at least a thin, faint band of weak signal grain -- or even outright mistracking sparkles -- appearing somewhere in the frame, especially in SLP. To hide that issue, auto-tracking circuitry within most VCRs worked to adjust playback tracking so that that band of grain/sparkles got positioned over the top of the VBI, keeping it out of sight. Fortunately, doing this didn't affect the vertical locking capabilities of late '70s+ era TVs, considering that their circuitry was already designed to cope with the heavy gaussian type noise that weak RF reception often put inside sync pulses anyway. (Only total signal dropouts, like if head switching were done in the VBI itself, was sufficient to confuse such TVs' sync circuits enough to make their pictures periodically jump or roll vertically.)
In any case, when you have gaussian grain or even "sparkles" randomly appearing over the tops of all those little white 1s and 0s making up your VBI teletext, XDS, and CC, you end up with unreadable teletext, XDS, and CC.
The only thing I can't explain is why some people experience success with recovering CC but not teletext. Could the visual teletext bits be finer/smaller in size than the CC and XDS bits? If so, the low luminance bandwidth of formats like VHS could just be smudging them, and them alone, beyond readability.