My PhD thesis addressed a similar question. I did a survey of sub-domain sized fragments shared between different protein folds. It turns out that there are plenty, even among folds considered evolutionarily distant.