The fish, that really was just a fish.

"A particularly interesting case which illustrates both the problem
of convergence and the danger of judging overall biology on
skeletal grounds is that of the rhipidistian fishes. For nearly
a century these ancient lobe finned fishes, as they are often
known, have been generally considered to be ideal amphibian
ancestors and have been classed as intermediate between
fish and the terrestrial vertebrates. This judgment was based
on a number of skeletal features including the pattern of their
skull bones, the structure of their teeth and vertebral columns and e
ven the pattern of bones in their fins, in all of which they closely
resembled the earliest known amphibians. It was assumed that their
soft biology would be also transitional between that of typical fish
and amphibia. But in 1938 fishermen in the Indian Ocean, off Cape
Province in South Africa, hauled to the surface a living relative of
the ancient Rhipidistia - the coelacanth. It was an astonishing discovery,
as the coelacanth had been thought to be extinct for a hundred million
years. Because the coelacanth is a close relative of the Rhipidistia,
here at last was the opportunity to examine first hand the biology
of one of the classic evolutionary links. Its discovery provoked
considerable excitement. Peter Forey comments:

'We had to wait nearly one hundred years before discovery of the
Recent coelacanth. During that time many fossil coelacanths were
described and, on the basis of osteological features, their systematic
position as near relatives of the extinct rhipidistians and as tetrapod
cousins had become part of "evolutionary fact", perpetuated today
in textbooks. Great things were therefore expected from the study
of the soft anatomy and physiology of Latimeria. With due allowance
for the fact that Latimeria is a truly marine fish, it was expected
that some insight might be gained into the soft anatomy and physiology
of that most cherished group, the rhipidistians. Here, at last, was
a chance to glimpse the workings of a tetrapod ancestor. These
expectations were founded on two premises. First, that rhipidistians
are the nearest relatives of tetrapods and secondly, that
Latimeria is a rhipidistian derivative.'

But examination of the living coelacanth proved very disappointing.
Much of its soft anatomy, particularly that of the heart, intestine
and brain, was not what was expected of a tetrapod ancestor.
[....]
If the case of the coelacanth illustrates anything, it shows
how difficult it is to draw conclusions about the overall biology
of organisms from their skeletal remains alone. Because the
soft biology of extinct groups can never be known with any
certainty then obviously the status of even the most convincing
intermediates is bound to be insecure. The coelacanth represents
yet another instance where a newly discovered species, which
might have provided the elusive evidence of intermediacy so
long sought by evolutionary biology, ultimately proved to
be only another peripheral twig on the pre sumed tree of life."
(Evolution: A Theory In Crisis
By Michael Denton :179, 180)