>>In general, I get the feeling that these SMB2+ cache parameters are for situations where files are infrequently or rarely shared via SMB on a network. To be fair, how often are files shared other than flat-file/ISAM databases? Multiple people can open the same Office document but the 2nd and later users get it read-only, as sharing goes it's not sophisticated. In that scenario the default cache settings can reduce network traffic and load on the server.
>
>And that seems to be the only scenario they teach at M$ courses. As early as 1998. I had some networking issues (files closed at random) and the customer had a network guy come to check. He just opened a word document across the network, saved it and said everything was fine, it must be our app.
>
>Later it got worse.

I do recall a situation we had a customer site where the network guy was insisting that there was a bug in our software and not a network problem. I then proposed a simple experiment
1. Find a largish file - preferably something at least a few tends of megabytes in size and NOT related in any way with our software.
2. Perform a copy of the file -- If this file is network, copy it to a local drive, and if it is on the local drive, create a copy on the network.
3. Compare the contents of the original and network copy
He insisted this was irrelevant and a waste of time -- I responded with "just humor me, try it" I did notice that the copy operation seemed to take longer than I would've expected. The first comparison showed no differences -- to which he said "this proves nothing." So I asked him to perform the comparison operation again. This time there were some differences I noted -- to which I asked him if this result makes any sense. Apparently unconvinced, I had him try the comparison a few more times -- on some occasions no differences were detected, and other times one or more differences were detected (and the positions were different each time). I repeated my question "does these results make any sense?" Frustrated, he stated that this was a waste of time, that I obviously didn't know what I was doing or talking about. I then too a slightly different tack -- I asked if we copied a file from one location to another, should there be any differences between the original and the copy? He answered "obviously there shouldn't be any differences." I then asked, "if that's the case, then how do you explain the results of this experiment we just tried?"