Think less "50 kilometers" and more "50 meters". "Robotic teletanks" is much more in line with the technologies that have been actively demonstrated in laboratories rather "autonomous combat droids", which are based on technologies that said laboratories are using to hypebeast overly optimistic billionaires.
If we are talking about 50meter range, than it is just about pointless. That is the kind of number you have for weapon effects radius, let alone weapons range or maneuver range. That kind of numbers makes it only useful against utterly immobile opponents: aka mines and structures.
If one wants a vanguard in even low intensity conflict against current day rebels, one would need 4km range (root out range ATGM shooters). For protection against modern ground forces one would want to push out to 15km+ as thermals (on a mast) can see that far and drop fires.
----------------
In any case, without effective AI, a "radioman operated" teletank is just hopeless as all one manned tanks in history. If effective AI enables a tank to be operated with one man effectively, the "manned tank" could be operated with one man as well. (or at most two, if it has even far more weapons and sensors than the teletank)
A teletank is the UGV form factor with the least impact on combat operations. The concept may be easiest but the tech requirement (reliability/reaction time/etc) is actually quite hard. There are a lot of UGV concepts that has lower tech requirement and higher impact. If one is talking about urban combat, it is complex (terrain, tactical and even non-military human factors), demands good situation awareness and delicate use of force, all great difficulties for AI.
For a urban combat tank, a two man heavy tank (add AI and datalink to offload crew workload) with a compact weapon/ammo (no sabot pen, its not Anti-tank) can be all but passively RPG proof in all the common engagement angles, and add an APS and it'd take a near impossible tactical ambush to knock out by infantry or indoor emplacements. Short range combat actually favors armor quite a bit, as KE missiles are impossible, large warheads hard to handle in close combat under constant suppressive fire, and massed fire difficult with LOS obstructions.
In any case the infantry take far higher total losses than tanks and consumes more manpower. The kind of thing that have bigger impact are tasks replacing the infantry, replacing humans in the huge logistics tail or in tasks that was too suicidal for infantry.
In an urban assault role, a robot/sensor that can be the first into a door/window/hole can save a lot of lives and change how combat is fought altogether, if one could figure out the networking when one wants to pack a entire city full of them under enemy ewar.
For field combat, an cheap small semi-stealthy vanguard UGV that just moves sensors out forward, plugging gaps in aerial recon would be great. It can be cheap enough that getting some stuck is not a big problem, and would mostly serve as a anti-ambush tripwire that prevents mass casualties events.
Robotic logistics vehicles can immediately reduces the problem of drivers and can almost be off the shelf from the civilian world to boot. In this same line of development, robotic fire support vehicles is a natural addition at small additional development cost.
For offensive maneuver, in that other thread I was talking about UGVs that you could drop from penetrating air. A UGV that can hide and harass and fight behind enemy lines while being deployed by means as violent as being dropped out of bomber bays if not cruise missiles can generate novel problems for the enemy. Imagine standoff A2AD assets getting hunted down, or airfields getting attacked with loss of hugely important force multipliers.
In almost none of the above uses would the robotic vehicle be best commanded by the tanks. There is previous cases of control via infantry (EOD robots), local force command (breaching bots), or even centralized air force command (igloo white).
----------------------------
Even in the case of urban combat, the improvement in modern sensors and communications increasing means that the vehicle crew is increasingly better off handing decision making to higher levels. If you've seen videos of recent urban conflict, a commander watching UAV feeds of the battlespace and micro-managing combat elements directly (move to next corner, shoot that 3rd build, etc) is increasingly effective. A top down view (plus networked sensor fusion if available) enables far greater situation awareness and can know things like VBIED speeding down around the corner or an RPG team trying to flank via a back alley street. A tanker trying to watch a bunch of UAV feed and operate a tank at the same time is in for overload and do not have good situation awareness for controlling other forces. If a tank crew man has to get commanded from higher levels and operate a remote tank, that is just complication with little gain.
In any case, the main advantage in tele-tank operation is reduce manpower needs. How can there be reduced manpower if each tele-tank is tied to an operator? How operating savings happen is by having tanks outside of active combat (moving in rear areas, quite areas, etc) run without a full human monitoring, while maximize human effort on important combat zones. A tele-tank moving back to rearm on a road in safe areas probably gets no human supervision while a tank penetrating enemy defensive lines probably gets a dozen humans watching and tagging every bit of sensor output out of 360 degree array. This demands a networked and centralized control concept.
------------------------------
Some part of me do wonder if the real change to UGVs is VBIED spam becomes unbounded. Imagine a 2030 world where just about every newer motor vehicle have some self driving capability while easily carrying a ton of explosives out of any garage in any building in all those urban areas. Perhaps paving every road with anti-tank mines would be the first step in future urban campaigns....