Why would a limb end up with an odd number of toes?
Digit number is set during early limb development, and it can be reduced in more than one way. In mammals, evolution has trimmed toes many times, and the timing of that trimming differs between species: some lose digits early in limb formation, while others lose them later [3]. Because these are separate developmental events, there is no rule forcing the final count to be even.
The jerboa illustrates the evolved route. It is an odd-toed animal whose reduced digit count arose through evolutionary digit loss [3]. That makes it a real, naturally occurring case rather than a deformity—an animal whose normal body plan includes an uneven total of digits.
Uneven digits can also appear as developmental accidents
Not every uneven count is an evolved feature. A transgene insertional mouse mutant called "footless" shows asymmetric limb truncations, with digits missing on some limbs but not others [4]. The result is an animal whose total digit count is uneven, but as a consequence of a disruption in development rather than a species-typical trait.
This distinction matters for interpreting the question. An odd total can be a stable, inherited part of a species' anatomy, as in the jerboa, or a one-off outcome of a developmental error in an individual, as in the mutant mice [3][4]. Both are genuine cases of uneven digit totals, but they arise through very different routes.
What this evidence does not tell us
The available papers document specific cases rather than surveying how common uneven digit totals are across animals. The jerboa example comes from a study of the mechanisms of digit loss in mammals [3], and the mouse example comes from a study of a single mutant line [4]; neither is a broad census of digit counts.
It is also worth separating this question from a different one about counting ability. A review of number sense across the animal kingdom describes species that can assess the number of items in a set, but it does not address how many digits animals have [1]. Likewise, a book on digital technologies and human–animal relationships does not report on digit counts at all [2]. Those sources are about numerical competence and about animals in digital spaces, not about anatomy.
Sources used in this answer
Neuroethology of number sense across the animal kingdom
Reviews the neuroethology of number sense across many species, covering neural mechanisms of numerical competence rather than digit anatomy.
The Internet of Animals Human-Animal Relationships in the Digital Age
Examines how digital technologies are changing human–animal relationships, with no reporting on animals' digit counts.
Patterning and post-patterning modes of evolutionary digit loss in mammals
Shows that evolutionary digit loss in mammals can occur at different developmental stages, and identifies the jerboa as an odd-toed animal.
Asymmetric limb malformations in a new transgene insertional mutant, footless
Describes a transgene insertional mouse mutant with asymmetric limb truncations and missing digits, producing uneven digit counts.
Changes in the skeletal dimensions of the digits of the thoracic limb of Holstein cows after maturation
Studies growth-related changes in the skeletal dimensions of Holstein cow digits, without addressing uneven digit totals.