Hello Vincent,
Thanks for your explanation, pretty reasonable to use just a single link (more or less is exactly like IRF BFD). I asked because I was just wondering if creating a two interfaces LAG for the VSX Keepalive link would be, at first, possible and, if implemented, in case of VSX ISL links all down to have a more resilent heartbeat between VSX nodes would have helped (in any case, as suggested, a separate VRF specifically defined for VSX Keepalive should be used to keep it separated from anything could happen to VSX ISL).
Kind regards, Davide.