Thanks for the suggestion, however...
As noted above: the previous invocations of ceph osd pool create have exactly the same
structure (and succeeded) as the one which fails.
I used the long form because trying to use the short form
(ceph osd pool create ec2pool 2048 2048 erasure glasgow-eci-test, for example)
produced other "fatal" errors [complaining about my not specifying the final
numeric positional parameter, even though it's optional].
Weirdly, though, you're right: going back to the short form suddenly works again...
so I'd like to revise my question to:
"why is the behaviour (and requirements) of ceph osd pool create inconsistent -
sometimes failing if you don't specify all the parameters, even optional ones, and
sometimes failing if you do, despite the 'rules' you're following being the
same?"