How long should an hourly strategy last OOS?

What is a reasonable life expectancy of a strategy on hourly bars for Forex majors, in terms of IS and OOS periods?  For example, can I expect the strategy to behave properly in live trading for 50% of IS period?  (This means, it has remaining expectancy of  0.5 x IS – OOS)

This is a question that would deserve some serious research to give real answer.

In my opinion robust strategy is such that would work “forever”, like many daily trend following strategies – they are here for 40-50 years already and they continue to work.

Strategies on smaller timeframes would be more sensitive to actual market conditions, perhaps somebody else here would give his experience?

I think life expectancy is the most important factor. Based on our experience and common sense we should pick a number, lets say 1 month for 8 months old M15 data, and build the rest around it, trying to find settings which will give this 1 month steadily.

With optimization, I think that any expectation beyond 2 years life time for hourly FX strategy is unreasonable. Strategies generated over 8 months (or any IS period < 2 yrs) make me very uncomfortable.  I want future life expectancy of at least as much as OOS period.  There should be a way to detect when a particular instrument fundamentally changed its price behavior.  Based on monthly charts, ATR or some other long term metric that measures trendiness.  For example, monthly chart shows some turning point that spells trouble for an hourly strategy?

Eight months was just an example, personally I think number of trades is more reliable factor. What I mean is if you want results for a specific period, the best approach would be to include it in fitness algorithm.

Good and crucial question. Actually hard thinking how to detect moment when strategy fails. 

Working on this concept:http://en.wikipedia.org/wiki/Statistical_process_control

Any experience?