Sorry for the late reply..
The 1756-ENBT built-in diagnostic web pages (and the ones on the 1734-AENT as well) are the first place to look.
The -ENBT will give you an idea of the packets/second and the CPU loading. That will let you put some quantitative measurements on the loading of the module.
CPU loading is low (< 35%). Packets/second is 888 with 90 rejections (should I be concerned about the rejections)
Both web pages will also let you check for media error counters that might be increasing.
There are a lot of FCS & MAC transmit Errors
A fixed speed/duplex port is always a big red flag for me... I've seen a lot of duplex mismatches or autonegotiation failures that cause collisions or errors on the port. Why did you use fixed speed and duplex ?
Based on internal discussions & some reading. Mainly to prevent the autonegotiation going into an endless battle & not recovering.
Are you using managed switches ?
We are not using Cisco 37xx series in multi-stack config. Previous was 3com stack. Interesting here on the 3com, I could ping across the entire stack, but the ENBT could not see any IO/drives plugged into different stack modules. Therefore they had to be all on the same module. This drove itself to the replacement with the Cisco equipment.
Are you using unicast or multicast connections ? If multicast, have you made provisions to handle the IGMP group snooping and querying functions the switches might have ? I am pretty sure the Cisco gear has it on by default, but will double check.
Multicast as far as I know, think Unicast is firmware/version/model dependant??
Remember that there will be at least 3 different RPIs for the 1734-AENT module; one for the Adapter itself and the Rack Optimized connections to the discrete modules, and one each for the Analog modules.
ahh, wasn't as aware of that as I should of been.
But it did lead me to something.
For some reason (& again I think this was based on internal discussion), each module had a different RPI. There was a reason, but with some GSV traps that I put in place, I was getting a lot of re-connections (> 8/hr).
Except for 1 module that was at 50mS. So I have them all set to 50mS, & for the first 48 hr, there were no re-connections that I captured.
Now, 5 days later, the reconnections are only 1/day, at random times from each other.
We are also going to change the CPU from the L55 to a L61 soon, this has been on the cards for some time just have not had a break in production to do so.