Anytime a transaction has a timestamp attached to it, and there are signed but not submitted transactions at play, shouldn't the holder of the non-submitted tx have to take into account the confirmations w.r.t. the expiration time?
Namely in example 7 of this wiki page on contracts I would think that if the AP waits until nearly midnight (almost when the tx expires) then there is a possibility of a double spend and the AP may not get paid. That means the AP would have to submit the tx 6 confirmations early... almost one hour.
Is this a correct understanding of locktime in relation to double spend attacks?
How should this be handled when the network is rapidly increasing in block generation (and difficulty hasn't been adjusted)? This is happening now with the GPU-> ASIC conversion