I think it might be better to protect this in the guardian on the host side. We already have a mutex called submitDataLock that means L1 block and L2 batches are never submitted in parallel, perhaps we add submit tx to that same lock?

I just imagine that validateTx is read-only and if it's getting interfered with then it's probably clashing with one of those other methods, WDYT?

LGTM after discord discussion, sounds like we're only worried about the clashing for simultaneous validate tx and other data changes shouldn't impact.