It would seem that the fight for survival is between extensions of rationals rather than individual primes p. Intuition suggests that survivors tend to have maximal number of ramified primes. These number theoretical speciei can live in the same extension  to "cooperate".
Before starting one must clarify for myself some basic facts about extensions of rationals.
 Extension of rationals are defined by an irreducible polynomial with rational coefficients. The roots give n algebraic
numbers which can be used as a basis to generate the numbers of extension ast their rational linear combinations. Any number of extension can be expressed as a root of an irreducible polynomial. Physically it is is of interest, that in octonionic picture infinite number of octonionic polynomials gives rise to spacetime surface corresponding to the same extension of rationals.
 One can define the notion of integer for extension. A precise definition identifies the integers as ideals. Any integer of extension are defined as a root of a monic polynomials P(x)=x^{n}+p_{n1}x^{n1}+...+p_{0} with integer coefficients. In octonionic monic polynomials are subset of octonionic polynomials and it is not clear whether these polynomials could be all that is needed.
 By definition ramified primes divide the discriminant D of the extension defined as the product D=∏_{i≠ j} (r_{i}r_{j}) of differences of the roots of (irreducible) monic polynomial with integer coefficients defining the basis for the integers of extension. Discriminant has a geometric interpretation as volume squared for the fundamental domain of the lattice of integers of the extension so that at criticality this volume interpreted as padic number would become small for ramified primes an vanish in O(p) approximation. The extension is defined by a polynomial with rational coefficients and integers of extension are defined by monic polynomials with roots in the extension: this is not of course true for all monic polynomials polynomial (see this).
 The scaling of the n1tuple of coefficients (p_{n1},.....,p_{1}) to (ap_{n1},a^{2}p_{n1}.....,a^{n}p_{0}) scales the roots by a: x_{n}→ ax_{n}. If a is rational, the extension of rationals is not affected. In the case of monic polynomials this is true for integers k. This gives rational multiples of given root.
One can decompose the parameter space for monic polynomials to subsets invariant under scalings by rational k≠ 0. Given subset can be labelled by a subset with vanishing coefficients {p_{ik}}. One can get rid of this degeneracy by fixing the first nonvanishing p_{nk} to a nonvanishing value, say 1. When the first nonvanishing p_{k} differs from p_{0}, integers label the polynomials giving rise to roots in the same extension. If only p_{0} is nonvanishing, only the scaling by powers k^{n} give rise to new polynomials and the number of polynomials giving rise to same extension is smaller than in other cases.
Remark: For octonionic polynomials the scaling symmetry changes the spacetime surface so that for generic polynomials the number of spacetime surfaces giving rise to fixed extension is larger than for the special kind polynomials.
Could one gain some understanding about ramified primes by starting from quantum criticality? The following argument is poor man's argument and I can only hope that my modest technical understanding of number theory does not spoil it.
 The basic idea is that for ramified primes the minimal monic polynomial with integer coefficients defining the basis for the integers of extension has multiple roots in O(p)=0 approximation, when p is ramified prime dividing the discriminant of the monic polynomial. Multiple roots in O(p)=0 approximation occur also for the irreducible polynomial defining the extension of rationals. This would correspond approximate quantum criticality in some padic sectors of adelic physics.
 When 2 roots for an irreducible rational polynomial coincide, the criticality is exact: this is true for polynomials of rationals, reals, and all padic number fields. One could use this property to construct polynomials with given primes as ramified primes. Assume that the extension allows an irreducible olynomial having decomposition into a product of monomials =xr_{i} associated with roots and two roots r_{1} and r_{2} are identical: r_{1}=r_{2} so that irreducibility is lost.
The deformation of the degenerate roots of an irreducible polynomial giving rise to the extension of rationals in an analogous manner gives rise to a degeneracy in O(p)=0 approximation. The degenerate root r_{1}=r_{2} can be scaled in such a manner that the deformation r_{2}=r_{1}(1+q)), q=m/n=O(p) is small also in real sense by selecting n>>m.
If the polynomial with rational coefficients gives rise to degenerate roots, same must happen also for monic polynomials. Deform the monic polynomial by changing (r_{1},r_{2}=r_{1}) to (r_{1},r_{1}(1+r)), where integer r has decomposition r=∏_{i}p^{ki} to powers of prime. In O(p)=0 approximation the roots r_{1} and r_{2} of the monic polynomial are still degenerate so that p_{i} represent ramified primes.
If the number of p_{i} is large, one has high degree of ramification perhaps favored by padic evolution as increase of number theoretic cooperation. On the other hand, large padic primes are expected to correspond to high evolutionary level. Is there a competition between large ramified primes and number of ramified primes? Large h_{eff}/h_{0}=n in turn favors large dimension n for extension.
 The condition that two roots of a polynomial coincide means that both polynomial P(x) and its derivative dP/dx vanish at the roots. Polynomial P(x)= x^{n} +p_{n1}x^{n1}+..p_{0} is parameterized by the coefficients which are rationals (integers) for irreducible (monic) polynomials. n1tuple of coefficients (p_{n1},.....,p_{0}) defines parameter space for the polynomials. The criticality condition holds true at integer points n1D surface of this parameter space analogous to cognitive representation.
The condition that critical points correspond to rational (integer) values of parameters gives an additional condition selecting from the boundary a discrete set of points allowing ramification. Therefore there are strong conditions on the occurrence of ramification and only very special monic polynomials are selected.
This suggests octonionic polynomials with rational or even integer coefficients, define strongly critical surfaces, whose padic deformations define padically critical surfaces defining an extension with ramified primes p. The condition that the number of rational critical points is nonvanishing or even large could be one prerequisite for number theoretical fitness.
 There is a connection to catastrophe theory, where criticality defines the boundary of the region of the parameter space in which discontinuous catastrophic change can take place as replacement of roots of P(x) with different root. Catastrophe theory involves polynomials P(x) and their roots as well as criticality. Cusp catastrophe is the simplest nontrivial example of catastrophe surface with P(x)= x^{4}/4axbx^{2}/2: in the interior of Vshaped curve in (a,b)plane there are 3 roots to dP(x)=0, at the curve 2 solutions, and outside it 1 solution. Note that now the parameterization is different from that proposed above. The reason is that in catastrophe theory diffeoinvariance is the basic motivation whereas in M^{8} there are highly unique octonionic preferred coordinates.
If padic length scale hypothesis holds true, primes near powers of 2, prime powers, in particular Mersenne primes should be ramified primes. Unfortunately, this picture does not allow to say anything about why ramified primes near power of 2 could be interesting. Could the appearance of ramified primes somehow relate to a mechanism in which p=2 as a ramified prime would precede other primes in the evolution. p=2 is indeed exceptional prime and also defines the smallest padic length scale.
For instance, could one have two roots a and a+2^{k} near to each other 2adically and could the deformation be small in the sense that it replaces 2^{k} with a product of primes near powers of 2: 2^{k} = ∏_{i} 2^{ki}→ ∏_{i}p_{i}, p_{i} near 2^{ki}? For the irreducible polynomial defining the extension of rationals, the deforming could be defined by a→ a+2^{k} could be replaced by a→ a+2^{k}/N such that 2^{k}/N is small also in real sense.
See the chapter TGD View about Coupling Constant Evolution or the article Trying to understand why ramified primes are so special physically.
