Not sure what you mean at "infinitesimal distance level".
With the caveat to follow, as one takes the wavelength of a photon towards zero (i.e., ultra high energy gammas), its contribution to the stress-energy tensor's expectation value at a point climbs. In turn the average radius of curvature around that point becomes smaller than the gamma's wavelength.
When we are in that regime measuring distances by wavenumber and measuring times by photon frequency gives useless results. Likewise, infrared photons with wavelengths comparable to the diameter of a black hole and finding themselves lurking around black holes, that is hard to understand too.
Now we introduce the foreshadowed clarification.
We can always find some ultraboosted observer who would (in its rest frame) measure that a photon emitted from your computer screen as an ultra high energy gamma ray. This doesn't mean your computer screen is creating small black holes. After all we can always find some other observer who sees a photon from your screen as having a wavelength of thousands of light-seconds.
So let's consider instead a photon scattering inelastically off a charged particle. The incoming photon momentum need not equal the outgoing photon momentum. Some of the momentum can be transferred to the charged particle ("recoil"), but some of the momentum can be absorbed into a composite charged particle's internal degrees of freedom. That might lead to, for example, https://en.wikipedia.org/wiki/Photodisintegration>.
But as we take the incoming photon energy ever higher, and deposit more of it into the charged particle, we start growing the stress-energy tensor's expectation value in the region of the scattering as above. The radius of curvature shrinks, and in extremis may keep daughter products (low energy outgoing radiation ranging from the outgoing scattered photon to the results of photonuclear and photosubnuclear processes; we can also get pair production from a high-energy charged particle inverse-compton-scattering a photon to sufficient energy) from leaving the immediate region.
What happens then? Semiclassical gravity, which is quantum field theory for matter on a classical curved spacetime background, gives us confusing results because the excitations in the quantum fields don't localize like a classical theory, and our averaging procedure (to obtain the expectation value) gives us a smeared stress-energy that less and less resembles the quantum mechanical system (and consequently the generated curvature looks unphysical).
The radius of curvature and the Compton wavelengths of the particles in the strongly curved region are small, but not infinitesmial.
"Possible Connection Between Gravitation and Fundamental Length ", Mead 1964 https://doi.org/10.1103/PhysRev.135.B849>, which you can access by pasting a Swedish scy hob prefix in front, takes the priority in the development of this line of argument. He finds that the characteristic length scale is on the order of 10^-35 metres. (cf. https://en.wikipedia.org/wiki/Planck_units>).
But we can have two 10^-35 m gammas pass right through each other without engaging this argument. The gammas do indeed generate stress-energy, but as noted before one can always find observers who will see these gammas as infrared, and their sky won't be full of black holes in one direction and devoid of any photonuclear interactions in the other. (We can even have extreme accelerators ("Unruh observers") who count multiple gammas where non-accelerated observers see just one -- in principle acceleration can make the observer's particle count climb arbitrarily high, but that doesn't result in the observed system becoming heavy, much less a black hole).
The wavelength (and particle number) in some appropriate inertial frame of reference for the scattering detailed above is what we want when analysing this type of interaction. This is usually the "COM" frame https://en.wikipedia.org/wiki/Center-of-momentum_frame#Speci...>, which can be calculated by arbitrary observers of our charged particle + photon (even ones who want to count more particles than an inertial observer, or redshift everything to very low energies).
tl;dr: What we're doing above is building a COM frame with a small spatial volume and high energy and asking what is the calculated curvature in that part of spacetime. That gets messy long before we have to bring in infinitesimals or start to worry about making a small black hole.