fix: Remove locks to fix testnet halt

[ValarDragon]

No description provided.

[ValarDragon]

See comments from me in prior issues, for why the race conditions are actually the race sensor being broken. Going to sleep now, so can't chase down links, but should be eventually linked to from here: #158

[ValarDragon]

Yikes, this fix does not maintain correctness in the generic case. (So would not be ok for the state machine long term, is fine to revive the testnet).

So need to do the deeper cache KV store data store correction

One note: currently we provide no guarantees about iterating during insertion... if you write to a btree (insert/delete), you invalidate all iterators currently iterating on that tree. I don't see that changing any time soon.
(source google/btree#28 (comment))

We deem this sufficient for testnet, so not going to merge this, but will instead make a testnet-only rc, as we make a better cache kv

[czarcas7ic]

Closing as I believe this is no longer needed