MCH1210 during DSPFD (QDBEXDME?)

I'm getting a weird error when doing DSPFD on a certain file.

The log sometimes looks like this:

dspfd <filename>
Receiver value too small to hold result.
Dump output directed to spooled file <n>, job <jobspec> created
on system <system> on 03/08/16 11:34:21.
Function check. MCH1210 unmonitored by QDBEXDME at statement *N,
instruction X'01C3'.

And sometimes looks like this:

dspfd <filename>
Receiver value too small to hold result.
Dump output directed to spooled file <n>, job <jobspec> created
on system <system> on 03/08/16 11:40:48.
Processing command failure, refer to job log for details.

In either case, when I issue the DSPFD command, there's a blip of a
status message at the bottom of the screen (I believe telling me it's
starting a dump), and then it shows me partial output, up to the
following:

Access Path Description
Access path . . . . . . . . . . . . . . . . : Arrival
Sort Sequence . . . . . . . . . . . . . . . : SRTSEQ *HEX
Language identifier . . . . . . . . . . . . : LANGID ENU

It doesn't show member description, etc.

I noticed while Googling that there were some PTFs that kinda-sorta
matched, like

http://www-01.ibm.com/support/docview.wss?uid=nas3028203848754b2e7862577120073bb37

MF49873 - OSP- MSGMCH1210-F/CBLABRANCH-T/QDBEXDME DSPFD GETS MCH1210

But my file does not have any LOB field. The other sort-of matches
that I found didn't seem to describe my situation either. Which I
realize doesn't rule them out as a fix for my situation. (And I'll
admit I didn't have the patience to wade through all of them.)

Some more interesting symptoms:

I had populated the file using an RPG program, which wrote about half
a million records. I could have sworn that before I ran the program, I
changed the size so that it could accept ten million records, but the
partial output I get on the problematic file shows

Member size SIZE
Initial number of records . . . . . . . . : 10000
Increment number of records . . . . . . . : 1000
Maximum number of increments . . . . . . : 3
Record capacity . . . . . . . . . . . . . . : 13000

which is still the default on our system.

I also recompiled the same DDS to create a file in another library,
changed the size of the new file to allow 10 million records, and then
copied all the records from the problematic file to the new file. Went
through fine. DSPFD shows half a million records in the new file.

Any ideas what is going on?

John Y.