This is a high-level answer.
- Much of the focus has been on nanotech as opposed to self-replication on the macro scale
Search of Google Scholar with keywords "self replicating machines molecular nano" reveals a slew of papers, especially in the early 2000's, with diminishment in activity subsequently. (My recollection is it turned out to be harder than initially hoped, but CRISPR has gotten a lot of attention in regard to molecular scale machines.)
There may be a weakening distinction between what constitutes a machine at this scale subsequent to CRISPR, i.e. organic machines vs. inorganic machines.
Biology provides numerous examples of self-replicating machines, but artificially engineering such complex systems remains a formidable challenge. In particular, although simple artificial self-replicating systems including wooden blocks magnetic systems modular robots5,6 and synthetic molecular systems have been devised, such kinematic self-replicators are rare compared with examples of theoretical cellular self-replication. One of the principal reasons for this is the amount of complexity that arises when you try to incorporate self-replication into a physical medium
Kim, J., Lee, J., Hamada, S. et al. Self-replication of DNA rings. Nature Nanotech 10, 528–533 (2015)
Programmable manufacturing systems capable of self-replication closely coupled with (and likewise capable of producing) energy conversion subsystems and environmental raw materials collection and processing subsystems (e.g. robotics) promise to revolutionize many aspects of technology and economy, particularly in conjunction with molecular manufacturing. The inherent ability of these technologies to self-amplify and scale offers vast advantages over conventional manufacturing paradigms, but if poorly designed or operated could pose unacceptable risks.
Rabani E.M., Perg L.A. (2019) Demonstrably Safe Self-replicating Manufacturing Systems. In: Schmorrow D., Fidopiastis C. (eds) Augmented Cognition. HCII 2019. Lecture Notes in Computer Science, vol 11580. Springer, Cham.
However, search using the terms "self replicating machines macro scale" does return many results, some recent.
This may be more what you're looking for:
This paper introduces the concept of a physical self-replicating machine for deployment on the Moon utilizing raw material available on the Moon. A detailed but selective review is given in order to highlight clearly the novel aspects of this concept. In particular, it is hypothesized that if electric motors and vacuum tubes can be 3D printed from the limited repertoire of lunar materials, 3D printing constitutes a universal construction mechanism. This follows from the observation that mechatronic components are the constituent parts of all robotic mechanisms. In particular, we examine the use of 3D printing of electronics as a physical instantiation of a Turing machine. Several general implications of such a self-replicator are considered including whether it constitutes artificial life and mitigation against runaway replication.
Alex Ellery; September 4–8, 2017. Building physical self-replicating machines. Proceedings of the ECAL 2017, the Fourteenth European Conference on Artificial Life. ECAL 2017, the Fourteenth European Conference on Artificial Life. Lyon, France. (pp. pp. 146-153).
Possibly this type of endeavor has more chance of serious funding, since the long-term rewards are so potentially great, and cost of using humans in those environments is presumed to be even greater.
(At least one super-power seems interested in a moon-base, and Elon wants to die on Mars;)
- Runaway self-replication seems to be a concern in all areas
This could be why there is some reticence (see gray goo—only needs to be "intelligent" at one thing;) but it would seem to be less of a concern at the macro-scale in a terrestrial setting at current level of AI. In this domain, it's almost certainly a question of cost vs. cost of human labor.