check_examples_run has no caller and fails open — and '-- --help' does not exit this workspace's examples #3261

check_examples_run has no caller and fails open

crates/aprender-orchestrate/src/stack/releaser_preflight/checks.rs:376 runs every discovered
example with cargo run --example <name> -- --help. Two defects.

1. Its only callers are its own tests. grep -rn 'check_examples_run\b' finds it referenced
only from releaser_preflight/tests.rs. Nothing invokes it against the real workspace at release
time — a facility with a self-test and nothing behind it.

2. It fails open.

// Runtime exit with non-zero is OK for --help
succeeded += 1;

A non-zero runtime exit counts as a pass unless stderr contains error[E or could not compile.
So an example that crashes passes; only one that fails to compile is reported.

And -- --help does not do what it assumes here

Measured on a random sample of five examples, 60 s each:

HANG     aprender-core::grid_search_tuning
HANG     aprender-serve::imp_700_realworld_verification
exit101  aprender-serve::debug_q4k_gemv
ok       aprender-simulate::reproducibility
HANG     aprender-serve::bench_barrier_overhead

Three of five ran past 60 s. These are not CLIs — they are compute demos that ignore argv and
do the work. --help does not make them exit. (Five is a small sample; the ratio is not the claim,
the kind is.)

So the checker, if it were ever called on the real workspace, would sit for 980 × its own timeout.

What is owed

1. Give it a caller, or delete it. A release preflight check that nothing runs is not a check.
2. Fix the fail-open: a crash is a failure.
3. Decide what "every cargo run --example" means for a workspace whose examples are demos —
   either "starts and does not crash within N seconds" with timeout treated as PASS (~2 h at
   N=10, so a nightly/full-surface job, not a fast lane), or a declared-runnable subset.

Context: found while building the T-1 deep lane (#3260). APR-RELEASE-001 §4 names
every cargo run --example as a T-1 clause, and #3260 implements the three clauses that are cheap
and unambiguous while leaving this one to a decision.

Refs #3260, #3239