Yes, `evalf/EllipticCPi`(nu,k) is buggy. E.g.

evalf(EllipticCPi(0.985, 0.4));                 # crash
plot(EllipticCPi(n, 0.91), n=0.97..0.99); # discontinuous

Obvious workaround: use  EllipticPi(nu, sqrt(-k^2 + 1))
You may redefine

`evalf/EllipticCPi` := (nu,k) -> `evalf/EllipticPi`(nu, sqrt(1-k^2));

f:=CurveFitting:-Spline(my_array, x, degree=1):
b:=0.7: a:=solve(f=b,x):
linevert:=plot([a,y,y=0.1..0.9]):
display(my_graph, line, linevert);

restart;
NumPartStrict:=proc(n::integer[8], m::integer[8])::integer[8];
  option autocompile;
  local k;
  if m=0 then return `if`(n=0,1,0) fi;
  if m>=n then return 1 + thisproc(n,n-1) fi;
  add(thisproc(n-m+k-1,m-k),k=1..m)
end:

NumPartStrict(15,8);
                               13
NumPartStrict(125,75);
                            3179161
NumPartStrict(150,50);
                            14682366
 

You want the complex integral, not the line integral.

restart;
f:=exp(I*z)/z; Z:=r*exp(I*t):
J:=Int(eval(f, z=Z) * diff(Z,t), t=0..Pi);
value(J) assuming r>0;
evalf(eval(%,r=1/2)) = evalf(eval(J,r=1/2)); #check

For a non-polynomial expression we must define a custom order and use an inert version. For this example:

f:=x^2+x^y+1-z:
`%+`(sort([op(f)], (u,v) -> y in indets(u))[]);
#        x^y - z + 1 + x^2

restart;
ex:=int(diff(u(x,y,t),x)*v(x,y,t)+diff(v(x,y,t),x)*u(x,y,t),x): 
simplify(eval(ex, v=F/u)): eval(%, F=u*v);
#                     u(x, y, t) v(x, y, t)

It is always better to use explicit (or parametric) plots when possible.

plots:-animate(plot, [[sqrt(x^3/(2*a - x)), -sqrt(x^3/(2*a - x))], x = -5 .. 5, view=-5..5, color=red], a=-5..5);

If you really want implicitplot, just add the option signchange = false.

You could use inert Sum instead of sum and then, when needed do e.g.:
value(eval(a, N=4));

This is obviously a cooked up example.
The minimum is 0, but it is attained not only for x=y=z=1,  but also for  any x=1, y=a, z=a  (a>=1).

int(sin(log(x+1)),x=-1..1, method=FTOC);

Should work in Maple 17 (the function has an elementary antiderivative).

So, you have

restart;
alias(a1 = a1(r), a2 = a2(r), a3 = a3(r));
#                           a1, a2, a3
array1 := [a1, a2, a3];
#                     array1 := [a1, a2, a3]
array2 := [seq(cat(a, i), i = 1 .. 3)];
#                    array2 := [a1, a2, a3]
array1 - array2;
#                             [0,0,0 ]
diff(array1, r),   diff(array2, r);
#              [ d       d       d    ]           
#              [--- a1, --- a2, --- a3], [0, 0, 0]
#              [ dr      dr      dr   ]           

To understand (and see) what happens, let's use macro instead of alias
 

restart;
macro(a1 = a1(r), a2 = a2(r), a3 = a3(r));
#                           a1, a2, a3
array1 := [a1, a2, a3];
#                array1 := [a1(r), a2(r), a3(r)]
array2 := [seq(cat(a, i), i = 1 .. 3)];
#                     array2 := [a1, a2, a3]
array1 - array2;
#            [-a1 + a1(r), -a2 + a2(r), -a3 + a3(r)]
diff(array1, r),  diff(array2, r);
#          [ d          d          d       ]           
#          [--- a1(r), --- a2(r), --- a3(r)], [0, 0, 0]
#          [ dr         dr         dr      ]           

So, as we see, array1 and array2 are totally distinct, array2 does not contain arguments.
The fact that array1 - array2  evaluates to  [0,0,0]  (for alias only!) is (most probably) due to (automatic?) simplification where the alias acts.

In general, when using alias (or macro), it is not wise to use "special" constructs such as your cat (involving aliased names). But at least, with macro, we see how Maple interprets our expressions.
 

An alternative for Carl's solution.

delind:=proc(T::table, S::set, M::set(list):={}) 
  local i,X; 
  unassign(subs(X=T, [seq(X[i],i=S),seq(X[i[]],i=M)])[])
end proc:

T:=table([11,22,33,44,55,66]): T[a,b]:=c: eval(T);
#           TABLE([1 = 11, 2 = 22, 3 = 33, 4 = 44, 5 = 55, 6 = 66, (a, b) = c ])

delind(T, {2,3});
delind(T, {1,5}, {[a,b],[6]});  eval(T);
#           TABLE([4 = 44])

Edit. @Carl. I think that you prefer:

delind:=proc(T::table, S::set, M::set(list):={}) 
  local X; 
  unassign(subs(X=T, `?[]`~(X,`[]`~(S)) union `?[]`~(X,M))[])
end proc:

Statistics is a modern package, full of modules and generated modules, not easy to debug.
For a workaround, as you noticed, it's possible to use (for the moment)

HR:=(X, t) -> (PDF(X, t)/(1 - CDF(X, t)));

HR(1/2*X1 + 1/2*X2, 0.4); 
        2.352941176

Use Tabulate(DF):  instead of  print(Tabulate(DF));
(notice the colon).