Encouraging smurf-amp measurements

This is encouraging news:

I ran a scan for smurf-amps of /24s in 200/8 through 210/8 in late
July that yielded 30,083 networks that returned duplicate
echo-replies. Today, I rescanned those networks, and found that of the
original 30k, only 21,255 returned duplicates -- that's a 29% drop.

Whether it means that more networks overall are properly configured, or
that different networks are misconfigured, I know not. Optimistically,
I'd like to guess that the former is true, but a complete scan would
be necessary to know.