Mistake Clinic

Module-Specific Mistake Radar

Symptom	Misunderstanding	Falsification
Isolated tasks vanish	Vertex domain inferred only from edges	Declare an extra vertex with no records
Zero-cost route disappears	Zero used as absent-edge marker	One zero-weight edge
Parent route uses the wrong channel	Vertex parent loses edge identity	Parallel edges with different weights
SCCs merge a one-way chain	Reachability confused with mutual reachability	Two cycles connected one way
Iterative DFS finish order is wrong	A stack alone was treated as recursive events	Branching graph with a back edge
All blocked tasks called cyclic	Kahn leftovers confused with cycle members	Cycle with an outgoing tail
Weighted distance fixed too early	Discovery mistaken for final extraction	Direct expensive edge and cheaper detour
All vertices marked unbounded	Negative-cycle reachability scope omitted	Unaffected branch and disconnected cycle
One pass claims only one edge	Snapshot invariant applied to in-place relaxation	Ordered two-edge chain
Tied MST edges all excluded	Strict and non-strict cycle rules confused	Equal-weight triangle
Flow stalls below optimum	Reverse residual capacity omitted	Matching reassignment fixture
Value looks right, witness impossible	Only scalar result checked	Corrupt one internal flow to break conservation
Assignment misses a possible pair	Maximal confused with maximum	Pairs (0,0),(0,1),(1,0)
Planner overpromises makespan	Worker constraints omitted	Two independent positive-duration tasks on one worker

Practice Mistake Checks

For P1–P8, inspect model, edge identity, isolated vertices and event semantics. For P9–P16, state weight assumptions, objective and some/every qualifiers before calculating. For P17–P20, verify conservation, residual pairing and both directions of reductions. In all cases, keep the smallest failing graph, violated invariant, repair and a fresh successful example. Do not invent mistakes to meet a quota.