The idea that a virtual particle doesn't have enough energy to become real is deeply misguided. It's imposing far too much reality on a virtual particle.
Consider the virtual photon in deep inelastic scattering (DIS):
It has a lot of energy. At SLAC it my be a 40 GeV beam scattering into the famed 8 GeV spectrometer. That's 32 GeV of energy going from electron to proton. What's the momentum? Well, it's more. For any real beam/scattered electron:
$$q^2=(k'-k)^2 < 0 $$
Of course there is a frame in which the energy is 0, and the square momentum is negative... that's just the nature of $t$-channel scattering.
What about $s$-channel annihilation? Also at SLAC's Z-factory, 40 GeV positrons and electrons were collided:
Now that virtual photon has $\nu=80\,$ GeV and zero momentum in the lab. It can't be real. It has the energy, but it's stationary, and light needs to go at $c$.
What is the energy of an electron in a vacuum polarization loop:
It's anything you want it to be. Evaluation of that in a scattering diagram, or the ground state of a hydrogen atom, requires integrating over all energies positive and negative (renormalization notwithstanding).
I know the idea comes from popularized quantum field theory where the virtual particle is always introduced with hand wave to the Heisenberg uncertainty principle....borrowing, popping etc.
This does more harm than good. It contradicts, or at least obscures, a lot of more important quantum concepts, such has: eigenstates are stationary states, they have no time varying observables. Also: the vacuum is Lorentz boost invariant. (Don't start me on the Higgs-as-molasses analogy). It's hard to imagine a bubbling sea of particles that is time invariant and boost invariant.
So when a state, e.g. and electron in a beam and proton in a liquid hydrogen target transition to a final state: an electron in the 8 GeV spectrometer and proton sprayed about end-station A (see the 1st diagram), the EM and quark/gluon fields take all possible configurations (coherently, and some not e.g., external radiative corrections), and an analytic solution is intractable.
The problem is solved by perturbation theory in which the field configuration/evolution is described by diagrams with virtual particles over ever increasing complexity. That is it.
Nevertheless, they can be very instructive. The DIS virtual photon has a longitudinal and transverse polarization defined by $k^{\mu}$ and $k'^{\mu}$ that affects how it scatters from the target. The kinematics of the $\gamma^*$ can tell you when you're scattering from charges or from magnetic moments.
If the beam or target is polarized, the virtual photon has circular polarization that behaves as expected with respect to scattering off spin initial and final states. (It can also interfere with $Z^0$-exchange as if we're talking about two paths on a laser table).
Virtual particle are very useful in that respect.
There is one fact that remotely resembles the "not enough energy" quip: as
$$k'^{\mu} \rightarrow k^{\mu}$$
(forward elastic scattering):
$$ q^2 \rightarrow m^2_{\gamma} = 0$$
which is no interaction. In the limit the particles don't interact, the exchange photon get closer to being "on-shell", or real.
