The question is if this is a test failure (in the case the tests are run with a broken shell), or (also?) a problem which can happen in updatedb (when /bin/sh is broken).

IMO we shouldn't care too much about broken (and ancient!) shells
- as you said already (it's reasonable to expect "echo" to work") -, and only prevent the test from resulting in a FP.

@James: do you think we should exempt such a system from this part of the test to avoid false positives?

On OpenIndiana 2019.04 (an illumos distribution) the tests/misc/help-version test of the GNU findutils 4.7.0 fails like this:

PATH=/export/home/newman/ws/oi-userland/components/sysutils/findutils/findutils-4.7.0/xargs:/export/home/newman/ws/oi-userland/components/sysutils/findutils/findutils-4.7.0/locate:/export/home/newman/ws/oi-userland/components/sysutils/findutils/findutils-4.7.0/find:/export/home/newman/ws/oi-userland/components/sysutils/findutils/build/amd64/find:/export/home/newman/ws/oi-userland/components/sysutils/findutils/build/amd64/locate:/export/home/newman/ws/oi-userland/components/sysutils/findutils/build/amd64/xargs:/usr/xpg4/bin:/usr/gnu/bin:/usr/bin:/usr/sbin:/usr/perl5/bin:/sbin
++ env find --version
++ sed -n '1s/.* p;q'
++ echo find
++ sed 's/$//'
++ echo oldfind
++ sed 's/$//'
++ echo xargs
++ sed 's/$//'
++ echo frcode
++ sed 's/$//'
++ echo locate
++ sed 's/$//'
++ echo updatedb
++ sed 's/$//'
* updatedb: bad exit status '0' (expected 1),
  with --help or --version output redirected to /dev/full
FAIL tests/misc/help-version.sh (exit status: 1)

We run the test suite with GNU utilities (sed aet al.) taking precedence in $PATH.

We don't ship locate/updatedb from GNU findutils (we deliver those from mlocate package), so, perhaps, they are not even meant to work.