This is only a guess, but I think that in the case of automated fast-acting arbitrage between two exchanges, one large and fast moving; and one small and slow then the result would not be price lags, but rather larger spreads on the smaller exchange.
This is usually the case.
I'd add that the larger exchange will always lead the price change for two reasons:
1. they have a higher trade volume (more liquidity)
2. spreads are lower (which means better prices all round)
As you can see both reasons self-reinforce each other. Ie. the lower the spreads, the more reason people would want to trade there; the more people trade there, the lower the spreads will be.
Which leads on nicely to the next question:
So you would confirm that making a big buy in TH exchange would rise the price in the MTgox giving the Bid in TH will soon be higher than the ask of MT but then the bot will pay the ask in MT and sell at the higher bid price in TH resulting on a clearer upward trend than the one doable with the same cash amount in MT. Of course the amount to buy at a lower ask price in MT might be enough as to lower inmediately the price in TH. But when would this kind of strategy be usefull?
Yes, thats right, if somebody placed a large buy (or sell for that matter) on TH, then TH will lead the price movement, however why would somebody place a large buy order on a less liquid exchange when the price is worse?
It does happen occasionally, in which case, anybody who actively watches BTC prices closely or those who arb will have some very easy profit. This will have the affect of adjusting the market prices accordingly, in the way that you mentioned.
When would this kind of strategy be useful? Well either to arbitrage, or if you know the order is large enough to move the prices on the lagging exchange (MtGox in your example), then you can pre-empt the price movement... maybe if you combine it with a geared order from bitcoinica then profits to you!