The jury is still out on temporary dyslexia, but apparently astronauts can experience it on a temporary basis on return to earth.
However, mixing up (anagramming) digits or letters is not dyslexia. And this is confirmed from another source:
A common myth closely associates dyslexia with mirror writing and reading letters or words backwards.
At least when done with letters/words, e.g. reading smile as slime, this type of anagramming error is (sometimes) called a "migration error" (or more elaborately, transposition of letters within words). Furthermore, doing this on a regular basis is called Letter position dyslexia (LPD), a peripheral dyslexia. Some people with dyslexia also exhibit LPD, but LPD alone is not widely reported, no doubt in part because LPD is not that well studied: the first study of the developmental form appeared in 2007, while the first study of the acquired form came out in 2001, as far as I can tell. The only known cases of acquired LPD are following brain damage.
People with developmental LPD (but not full dyslexia) fail to be identified as such in standard tests of reading in English because typical reading tests don't have enough "migratable" words, c.f. a 2012 paper that was the first to look at LPD in English (with just 3 cases though). Also of note, one of the three cases discussed there was also officially diagnosed with attention deficit disorder and another one was suspected.
One interesting aspect is that LPD is not caused by swapping phonemes; for instance someone with LPD can read cloud instead of could (even though there's no ell heard in could). Based on this it seems plausible to me it could affect number reading as well, but I haven't found any published research on migrating digits in numbers. A 2nd (2014) study of LPD in English included a verbal digit span test, but the LPD children did not have any problems with this test; they were actually above average on verbal digit span. The paper does not report any number reading tests, unfortunately.
