Two separate IO devices on separate ports. both have tags configured and
displayed on a page. If one device is disconnected its tag displays #COM as per
normal. However the tag from the other IOdevice, its data does not update for a
short while (stays at the last value). It then resumes normal updates.

The time that the data does not update for is related to the driver retry / timeout settings.
ie with the defaults you hardly notice it. If you extend the timeout / retries
then  the pause can be significant.

Limit of issue:

Through testing it was found that the problem is specific to the page level only. If the two tags in the test were trended.
While on the graphics page you can see that the data pauses when one IOdevice goes offline.
Looking at the trend you can see that it has been able to read the correct values during the time that the graphics page was showing paused data.

Also a seperate page (on another display client for example) which is only showing the tag that is still online, does not experience any pause.

Reason:

This behaviour is by design. The Citect client waits for all IOServer
responses to data reads that are required by the page (or times out according
to [LAN]retry and [LAN]timeout) before issuing further requests for data reads.
LAN parameters can be lowered to obtain a faster update rate on the good data
points, but if the driver has higher timeout/retry settings there may still be
a noticeable lag before the driver eventually times out. Also #COM may be
displayed prematurely if set too low.

Please note that although it has not been specifically tested, V7 will likely have different behaviour in this area due to Publish / Subcribe data interface between clients and server.