We were able to workaround #1221 (Virtual keyboard dismisses when focux moves from image/video caption to paragraph) by rendering, but not displaying, captions that should not be visible to the user (PR). This is really just a workaround though, and we should determine what the root cause of this issue is so we can address that.

Additional Information

1. The issue does not occur when switching to a PlainText-based component (i.e., code block) instead of a RichText-based component.
2. The issue does not occur if we remove the logic to not render captions that should not be visible to the user (this is the fix that was applied in gutenberg#16759
3. The issue does not occur if we remove the call to props.onSelectionChange(undefined, undefined) in RichText.

Point 3 seems promising, but I was not able to finish investigating it because I needed to divert to other issues for the upcoming Open Beta.