One problem of the TH-9800 quad bander is failure of the Transmit Driver FET (Q409), mine has rd07mvs1b
which in the spec sheets shows VDD at 7.2v for 7W out for test cct, but also quotes running at 9.2v,
however according to forums, this FET spec'd to run at 12v DC maximum,  however, checking the Mitsubishi Datasheet,
says VDSS max is 30v and VGSS max is 20v, so we are well under the max. supply voltage.
So running it at 13.8V is meant to be stressing it and if you run it up near 15v then it is considered will
definitely fail, then the theory from Datasheet means, it must be spikes above 30v to be killing it,
or  the heat dissipation is not good enough for the power it handling at 13.8v, which I estimate at about 15-16w,
this would seem to be the real reason for failure, not the fact of running at 12-15v. The max power output
from this MOSFET is 50w, but that is under basis of it being clamped to a big heatsink and probably with
help from a fan. As mounted in this radio, it not mounted on a heatsink, so power rating has to be considerably
lowered for this little MOSFET to dissipate the heat.

For this idea to run it at lower voltage supply to me says we really trying to low the power generated in this Driver stage.
The use of the 7808 regulator is simply easy, (or you could run a zener diode and resistor and transistor, as
long as you do the calculations for the desired voltage you want to run the FET at, i.e. 10v,)
because the lower the supply voltage at the FET dictates the drive power to the PA FET, so running at 8v supply
will lose a few watts from what will appear at the output, but at least you know the 7808 is stable and provides protection.

Alternative idea (not tried yet) is to replicate the cct for the Q64, which is a PNP 2SC1132 transistor, use a higher
rated transistor, something that can do a couple amps, a PNP equivalent to TIP31, and connect it in series with L660
as we did with the 7808, with Emitter on the supply side and collector on the FET side passing thru L660 to feed the FET.
This way the new transistor has its Base connected to Base of Q64, so that supply is only provided to the Driver FET
when you transmit, this does not stop the over heated damage, in fact it could increase the chance of failure,
as we now switching current on/off and the transients may be enough to kill the Driver FET, however we could use
this in combination of output of 7808,(or output of a zener/resistor/transistor) to provde Tx switched  supply to
the Driver FET, which is how it should have been done in the first place.

Summary:
IMHO, it not a supply spike problem causing failures, but the ability for Q409 to dissipate the heat for the given power
output it is running at, so the thought here is to lower the supply voltage to lower the power output of the Drive FET (Q409).
A few different approaches to fixing the potential problem and protect it, just a matter of how much effort required
to do it properly, or how much power losses you prepared to live with, if doing it the basic way. Or find a better
way to get the heat dissipated from this Driver FET.
In fact,considering how hot the case is when using this radio, you better off buying a 12v DC 5inch fan and glueing it
to the case and run continuously when radio is switch on, or add a simple switching cct, so fan runs slow on Rcv (~8v)
and full speed on TX (13.8v) so that you dont burn out the fan.