Isn't signing a license agreement for new contributions overkill? Most
other open source projects require nothing like that - just an
electronic text clearly stating the license that the author is using is
often enough.

For relicensing previously released code things are entirely different,
of course. But even in that case my impression is that people are being
unreasonable cautious when suggesting that any code that ends up being
rewritten should go through a full "clean team" process. The first time
I saw that use was by Compaq for reverse engineering the BIOS in the IBM
PC. They had left Texas Instruments which had refused to build a clone
for fear of lawsuits from IBM and so had every reason to be as careful
as possible. IBM did sue three clone makers in Brazil (1983) who copied
the BIOS directly but dropped the suit after they showed a token effort
to rewrite it themselves (basically just shuffling around non critical
code). This proved that Compaq had worried too much.

About the relicensing discussion in general, it is unfortunate that our
technical means (the author initials) don't match the legal reality.
Whoever wrote the original method from scratch is the author. If someone
made some change then they created a derived work and their copyrights
only cover the part that they changed. If we have permission to
relicense from the first author but not from the second then all we have
to rewrite is the small modification, not the whole thing.

About contacting the heirs of authors who passed away, I know from
experience that some will be very happy about it while others will be
extremely annoyed. You can't know which is which until you have tried.

-- Jecel