Yes, Thomson Reuters RIC codes are case-sensitive.
I believe the confusion comes from the fact that some frontends will 'assist' you if you misspell and automatically convert. That's just a trick. This doesn't change the fact that at the API level it is case-sensitive. In other words: if you are doing your own application you should definitely think of RIC codes as being case-sensitive. Always.
Side track: If you take your data from a Thomson Reuters market data system you basically request data via a data item name, not a RIC per se. This is because such a system is data source agnostic. A data item name is just a case-sensitive identifier for a particular data item. RIC is a particular naming system for data item names. A feed will expose its data items via data item names (duh!). In case of a Thomson Reuters feed they will always use the RIC nomenclature but other feeds on a Thomson Reuters market data system may use any nomenclature they like as long as it 'fits' inside the definition of data item name. For example on a Thomson Reuters market data system you may have a Bloomberg feed and such feed is likely to use Bloomberg identifiers as their data item names. So RIC is really a naming scheme and nothing more. It is not a technical thing.
I don't think there's a document that will state that RICs are in fact case-sensitive. It's just implied because nowhere is it said that they are case-insensitive.
Hope this helps.