Strange change of behavior in _Rreadf/_Rreadn, from V6 to V7

I've just noticed a strange phenomenon:

Calling _Rreadf and _Rreadn from C, intentionally reading less than a full record, specifying a buffer length of less than a full record (it was a full record a few years ago, but . . . ) has been working perfectly, throwing no exceptions, leaving nothing in the joblog, from V4 through V6.

Now, on this one V7 customer box, it's leaving one of these in the joblog for every such read (and since that box needs to read an entire 4000+ record file, it wraps the joblog):

C2M3003 Escape 20 02/09/12 17:10:56.811597 QC2IO QSYS . . .
From module . . . . . . . . : QC2RIODB
From procedure . . . . . . : C2M_ERROR_proc
Statement . . . . . . . . . : 4
To module . . . . . . . . . : WTCPVERIFY
To procedure . . . . . . . : main
Statement . . . . . . . . . : 123 *PRCLT
Message . . . . : Data was truncated on an input, output or update
operation. Cause . . . . . : 1. You attempted to read a record that is longer than the
length of the buffer you are reading into. 2. You attempted to write to or
update a record that is longer than the length of the record format you are
using. Recovery . . . : Use the correct record length when performing the
I/O operations. Technical description . . . . . . . . : The value of errno
is set to ETRUNC.

Anybody have any idea why this would suddenly start happening at V7? Anybody ever seen anything like it?

--
JHHL