Now existing splice design has two sets of commitment txes, one for pre-splice and one for post-splice. This is fine if we consider splicing of a single channel in isolation. But it is not fine if we consider splicing of multiple channels in a single batched transaction (for example, if some node wants to splice-out funds from one channel and splice-in to another). In that case, a theft attempt of one channel (i.e. publishing a revoked commitment tx) may prevent the splice transaction from ever confirming for the other channel. A theft attempt may be costless if e.g. there is no reserve or the channel happens to have been single-funded opened and the attacker never received funds via that channel (which is a good reason for the victim to splice-out their funds from that channel).