it was likely the black box.
in the old school, they would formula it out.
weight of vehicle, length of skid marks, impact damage, etc.
although
unless there's a full on proper investigation, they're wrong.
my girlfriend was in a crash once, the cops said they hit that fence at
80 MPH.
hmmmmm...
1982 K10 @ 80mph + fence
seems like they should have done more than just knock the fence over
from 80 miles per hour.  i don't care what anyone says, 80 is fast,
technically speaking, and that truck is very heavy.  also considering
that not wearing a seatbelt in a 30mph crash has been known to cause
death.  oh and plus no one was wearing seatbelts.
so sometimes they just throw numbers around, i think they think they're
cooler when they do that sh.t.

Software is not necedssarily needed, speedometer needles have stuck at speed
in accidents many years before onboard computers were installed on cars.
Roy

Hopefully as a mere matter of getting a ticket -- actually stuff the
car at 90 with no belt, "you should have such problems."

To get back to the original poster's question:  I doubt that they
would find  a speedo   stuck at 90 (in the car, that is;  if your
Speedo  gets stuck at 90; well,  you and all your buddies at the
senior center should have such problems).  Needles getting frozen at
some reading is an ancient theme from fictional and apparently some
factual accounts of aircraft crashes, violent disruptions of one's
person in the era of mechanical wrist watches, etc.    I am thinking
that that takes extreme g's, especially in the case of an instrument
where the position of the needle is a dynamic rather than a static
affair (stopping a mechanical clock might be a lot easier to arrange,
John Cameron Swayze notwithstanding).

A quick glance through the literature of a field not my own indicates
that a slap mark on the *face* of a speedometer is what investigators
are usually looking for after a motor vehicle crash, not the needle's
actually being frozen.    Mythbusters could doubtless have a lot of
fun with some old jalopies and an abandoned runway in this regard.

As others have pointed out, there are several other ways of
determining the crash speed through forensics at the general crime
scene, perhaps onboard computers, and occupants coming clean about the
situation.

Cheers,
--Joe