Yes I agree, one system (not done by me), the supplier who did the system assumed the Scada system was accessable from the plant by the operators (as per an existing system), however, the management were not keen on this, the supplier had to do a quick re-think & decided to implement a work around, the scada would reside in the control room (un-attended) appart from maintenance.
Product development produced or updated recipes in an existing IBM based system, this was ported to an SQL DB on a server, then using RDS to upload the the recipe via the Scada to the master PLC, this then copied it to the relevant batching PLC & uploaded to the HMI's own recipe system (the HMI's had no connection to the network). It was a bit of a botch but worked, unfortunately I was not involved with this side so did not know about it until commissioning started, it worked well, if the Scada was not availlable it meant it could be still used, there was some checking of recipes were current that everytime a recipe was selected, the batch PLC would send a message to the scada, that in turn interrogated the Database to verify the recipe eas valid.
When we upgraded the other system I insisted that the scada with two stations be in the production area, the system was upgraded to use the same recipe database as the other, to ensure that if the link between the scada & the SQL server was lost, every time the recipe was selected by the operator it was written to the local scada HD as a csv file, if the SQL server could not be reached, it would load the csv, there were no problems with syncronistation in reallity as every time the recipe was downloaded a copy in csv format would overrite the one on the local HD (note: not a small recipe structure it contained over 2,500 tags). Although the two plants worked in diiferent ways, the recipe originally designed for the newer plant had a recipe based on a 100kg batch (minimum size that could be handled), so the ingredient additions were based on that, when the operator selected a recipe they entered the batch size required, simple maths calculated each ingredient addition required, also extra fields had been added to each recipe stage for the new plant so these could be used for any special requirements for the upgraded plant, each recipe stage contained the information for the type of operation for example a field for the type of operation i.e. oil addition, water addition starch addition & so on, a temperature, if blending required so when a stage was loaded, it would contain all information but only the fields required would be used, even down to the transfer stage so for an ingredient addition only the fields required would be used, the others would be null or zero, the last stage i.e. transfer all fields for other operations would be null & only the transfer parameters used. This way we could have two plants that worked in different ways but use the same recipe.