Would a packet collision cause this on the plant wide intranet/internet
The nature of how TCP works means that it is not really possible for packet loss, packet collision (is this the 90s?) or other network 'interference' to cause you to read a one as a zero.
I think it is far more likely that your script (or maybe some sneaky code in your PLC) is setting the bit to zero.
I am having to go and "set" my 1's (reset) as they occasionally just zero off for no apparent reason
So Mango can write to these points too? Is that necessary for your plant to function? Is it possible to make the point read-only and see if the issue still occurs?