Possible work around for consensus issues

[mystenmark]

This PR is intended as a discussion topic, not as mergeable code. If we decide to go with this approach, we should take a slightly more principled approach in which we first determine that shared object locks are missing, and log somewhat loudly about the fact that we are proceeding without them.

[gdanezis]

Ack. Reflecting on it.

[lxfind]

Assuming this is on validators (not fullnodes), how does it ensure tx are executed in the right order?

[mystenmark]

Assuming this is on validators (not fullnodes), how does it ensure tx are executed in the right order?

The shared locks are still respected - an out-of-order attempt will just fail with lock errors. (But can be retried later once the parent certs have been executed).