Rapid-fire questions an interviewer asks around Functional Coverage Puzzles. Answer out loud, then check yourself.
Answer: Confirm scored denominator construction: which bins are legal, ignored, illegal, default, and thresholded by at_least. Wrong denominator means misleading percentages.
Trap: Jumping directly to percentages without verifying which bins are actually scored.
Answer: Candidates forget these filters reshape tuple space. You must compute filtered scored bins first, then map sampled tuples to that reduced denominator.
Trap: Using raw Cartesian product math even after explicit binsof/intersect filtering.
Answer: get_coverage reports the type-level number and get_inst_coverage one instance. With type_option.merge_instances = 1 the type number merges the bins of all instances (a bin hit in any instance counts); with 0 it is the weighted average of instance coverages. Use both so a weak instance is not hidden by a strong aggregate.
Trap: Treating a high aggregate as proof that every instance is healthy.
Answer: First prove reachability, then verify sampling timing/region, then inspect option thresholds and weights, and only then scale stimulus.
Trap: Adding random tests first instead of checking for unreachable bins or timing-model bugs.