Problems identifying text in Adobe Capture

It does the same on mine unless the font is super straightforward. It dos it thing on the fly, using resources from the cloud so there are chances that it stabilises if you wait a bit before scrolling (that's my experience). I also found that good lighting makes good contrast, and good contrast makes better identification.

Note that it suggests an average match based on fonts available on Adobe Fonts (Not ALL fonts) so you can be restricted.
I find it a good tool, "on the go" but worse than using a font identificator (might it be the one found directly on the Adobe Fonts, website or other ones like "what's the font" or "what font is"