A test of the hardness statements end to end, on a registered result. Elffers and de Weerdt prove single-machine scheduling with release times and deadlines NP-complete when the job lengths lie in $\{p, q\}$ for fixed $p > q > 1$ (formalized in lax-391470, on binary words). An instance with integer release times $r_j$ and deadlines $d_j$, all shifted up by the largest absolute value among them, is an instance of $1 \mid r_j \mid L_{\max}$ with due dates the shifted deadlines, and it is feasible exactly when the maximum lateness can be made at most $0$; its standard word with threshold $0$ is computed from the binary word in polynomial time.

Theorem. $1 \mid r_j \mid L_{\max}$ is NP-hard on the standard reading.

A test of the algorithmic side end to end: Smith's rule. On one machine, the total weighted completion time is smallest when the jobs run without idle time in the order of $p_j / w_j$ (jobs of weight $0$ last). A program reads the standard word of an instance of $1 \mid\mid \sum w_jC_j$ (not a problem of the zoo's data, which leaves out problems this easy) and its threshold, sorts the jobs by comparing $p_iw_j$ with $p_jw_i$, adds up the weighted completion times and compares with the threshold.

Theorem. $1 \mid\mid \sum w_jC_j$ is decided within $O(n \log n)$ instructions, plus the length of the word, on short words.