axis go to emergency after 1 hour of work

I try to be a week, the source of the problem but can not find. I have a gantry with linuxcnc 2.6 ubuntu 10.04, 2 parallel port, 4 brushless motor drive in step+dir.
Of course I have followed all the precautions for wiring ..... only limit swich not have the resistance of push -pull .... but so far everything is fine ( the machine works from 6 months ). the limit switch is on primary parallel port (on pc motherboard ...intel d255).
All work fine but every 1 hour axis go to emergency stop!!!

I try to connect and disconnect every emergency and chrge-pump switch .... but nothing change.

Normal and standard ubuntu 10.04 crontab action can do this (crontab of linuxcnc installation)?

I have touch screen with evede driver.

no message only my custom hal messages and limit switch messages.... sometime appear limit switch alarm ..... but I think it is false ..... (for 4-5 time no messages.... 1 time limit switch... some time switch 1 sometime 2 sometime 3 ..... but it is no real and I have debounce in every limit switch) ..... today I have see "voltage to ground" alarm in one drive for the first time .... a few hours after the protection fuse inside the drive is break .... i think i have some kind of discharges to ground.

thanks andy-pugh for the interest

I update the post with the cause and the solution of discharges to ground as soon as I will find.

I you have some other ideas, I listen to all the advice.

Is no exactly 1 hour .... save screen an other in bios is disabled ....
my dmesg ... example:
In my dsmeg I do not see things done wrong .... you will see some of them? I only see these errors due to the limit switch, but are dependent error of grounding ....

I add emi and emc filter in L/N power of pc .... in second time I think I add some filter or isolation transformer for the power part .... today not see other solution ....

Maybe debounce all the pins that could place the machine in emergency status.

5 day ago I've installed a surge arrester in the input 3F power .... these solve some strange problem .... than I checked all input pin whit hal scope .... I've found 2ms "0" random signal in pressure switch (normally my pressure signal is hight) ... I've debounce it ..... almost all emergencies disappeared ... now we have only 2 emergency event per day.

I believe that I will follow the advice of emcPT and put the debouncing of all signals that cause emergency for a time of 10ms....

N.B.: surge arrester He has taken away all that was not at all logical and without explanation.

Next day update