My own 2 cents addition...
Endothermic doesn't mean improbable. The motor of a reaction is not the loss of heat - since an other body has to absorb this heat - but something more subtle called entropy that tells how evenly energy spreads over the many degrees of freedom that can store it.
For instance, melting ice is endothermic, but melting is probable and even certain at +25°C.
There are situations where an object is expected to lose energy. A weight falls naturally. This is because its gravitation energy is concentrated in a single degree of freedom: height, which may store several J. Heat instead stores 2 to 4*10-21J per degree of freedom at room temperature, that is, each direction of movement of each atom, and similar things. Then, height converts into heat spontaneously when the object falls, as the J can spread more evenly into many degrees of freedom, while the opposite process doesn't happen.
So easily lost energy holds for mechanical energy, joules of electricity...
As opposed, it doesn't hold necessarily when the considered energy is already spread among many degrees of freedom. That's the case for heat, possibly for chemical energy - every time the unit energy change is comparable with R*T (for moles) or k*T (for molecules). Then, equilibria are possible, and are predicted numerically (or rather, properly extrapolated from a small number of measurements) through entropy, Gibb's energy, chemical potential, these things.
That's really a matter of energy concentration, not of energy nature. If you build sensitive electronics or accelerometers, you observe a residual fluctuating electrical or mechanical energy called a noise, that stores about kT/2 and won't disappear spontaneously.