I think most HFT systems have fairly smooth equity curves (ie. You wouldn't find a true HFT which performs vastly different from one day to another), which makes backtesting them a little unneeded when you can live test them with low risk and draw assumptions on long term profitability assuming market consistency, at which point you will have to reanalyze your system if it starts losing money in a new market.
Interesting... I think that a huge issue is that a lot of the prints reported on the tape are executed off exchange. So, there is no way that you could get filled on them. Best that I can tell from the SEC required reports, is that the broker really like to fill market orders, but tend to route limit orders to the exchange. So rather than pay the taker fee, they fill the order and collect the spread. I also belive that HFT's can work in your favor. They have a active "inside market" and a passive "outside market". That is a very good hint. I use simple stop orders for exits because I can model them, but maybe I should use limit orders even if I cant model the fills as well as I would like and just tweak it in live code. I know that I am leaving money on the table by letting my broker fill these at the market price.