The first sed example strips out (specifically) what is known to be not wanted - it substitutes "null" for it.
The latter case ensures only non-digits start the record, and looks for a string of digits - followed by anything (which could include digits). The bit in parentheses - the string of digits in this case - is retained by the back-reference "\1".
Obvious ain't it ....
. You have to be
real careful constructing regex so you don't get unintended results. We've all been there.