Firstly, I would look closely at your timing. What signal do you use to set the direction on the 74hc245?

What you describe can normally be atributed to not enough setup time for the HC374. Why only D7? maybe it has greater load or a different path. On your FRC connector maybe there is another signal coupling in on the D7.

Examine the route of your D7 signal.
Do a timing analysis. You have two major paths - the data path and the address path. Your data has to be stable for a few nS before the clock signal goes high on the hc374 - this is called the setup time (tsu on the data sheets). There is also the hold time(th), but in this instance you should meet this.

If you put another two inverters into the HC374's clock path does that make any difference? If it does, it almost certainly suggests you had a setup problem.

Another thing to try is to remove the inverter on the clock to the HC374 - clock the data in on the end of the bus cycle rather than the beginning.

The proper method is to do a timing analysis using worse case parameters and also measuring the timing use a logic analyser or oscilloscope. Using worse case parameters ensures your design will work over the whole temperature and process variation range thus having a reliable design (at least timing wise!!)



Lastly you might have a dodgy i.c