Fun with compiler upgrades

I have a fun story to share and a question.

The fun story is that during our upgrade, we had a period where our development machine was at 7.3 and our production machine was still at 7.1.  We would compile on our development machine with TGTRLS(V7R1M0) in order to install on production.

Well, now out machines are all at 7.3 and I removed the TGTRLS from our compile commands.  Everything was wonderful until today, when a program that had compiled just fine wouldn't compile anymore.  We were able to trace the event to the upgrade and then I tracked down the line that was failing.

Well, it was definitely a mistake on my part.  I had originally created a qualified externally described data structure but then changed it to unqualified.  No problem, except I forgot to remove the qualification on a couple of lines of code:

     saveMYFIELD = dsName.MYFIELD;

The fun bit is that those lines still compiled!  It's a large, complex program with lots of strange behavior in it, and something particular to the source allows that to happen, because if I try it with a shorter program (to try to isolate the issue) I can't get it to compile; I keep getting error RNF7030 saying "MYFIELD" is not defined.  It's only under a very specific (and currently undefined) set of conditions that this line of incorrect code compiles (and works!).

And whatever that edge condition was in the 7.1 version of the compiler, it's been fixed in 7.3.  Weirdly, I don't get the error I expect .  Instead, I get an unfamiliar RNF7591 error saying "An operand of a qualified name expression is not valid".  So those lines of code have to be fixed, which is perfectly fair.

The question I have is whether this sort of behavior - where IBM fixes a loophole - is documented somewhere.  I haven't been able to find it anywhere online, but I suspect I may have to dig in somewhere else.  If anybody knows where I can look, please let me know.

Thanks!