The scrambled words game is very useful in persuading the less sophisticated to take a passing interest in their own cognitive processes! it is intriguing and also rewarding as it shows we can do something apparently rather difficult more easily than expected. However, the difficulty will be greater for second language learners at earlier stages of study.
Thinking about the unscrambling process is also a useful illustration of certain established principles of cognitive psychology, many of which have a close relation to information-theoretic measures like message redundancy and channel noise. Such matters may well be further detailed in the references given by Leo, so here I will just mention a few obvious ones without going into subtleties. Those setting cryptology-based puzzles, like anagrams or crosswords, have probably given considerable thought to many of the same themes.
Most examples of word-scrambled texts I have seen conform to the case where individual words have the usual separators (spaces, commas etc.). It is a reasonable hypothesis that primary word-segregation is a (pre-conscious) analysis, fairly well automated.
Topical familiarity, general word-frequency
Another fairly automated effect, though quite dependent on individual experience
The edge-effect enhances perception of letters at the beginning and end of a word. Scrambled words with the first and last letters intact also greatly benefit because these strongly-perceived letters are precisely the ones often generating superior cues to the associative thesaurus, perhaps including a subvocal phonetic component.
Whilst short words are objectively easier, word-length itself provides significant information, and in longer words this may partially compensate for increased complexity.
Less familiar (or exotic) letters are more readily noticed in scrambled text due to an analogue in selective attention to effects of feature-contrast long known of in studies of short-term memory. Such letters assist also because they restrict search, due to their presence in a much smaller subset of lexical items.
This is dependent on the ongoing robustness of the perception of meaning. A text which would be challenging even unscrambled will offer much greater difficulty if scrambled.
Elimination via syntactic constraint
English, whose case-endings and verb conjugations have largely been smoothed away is particularly dependent on word-order in phrases as a cue to the function of a particular word (as verb, noun, adjective, adverb &c)
Internal morphemic articulation of words
Various issues partly encapsulated in the German example cited by Leo.
In longer words characteristic high-frequency constituents (things like "-ness", "-tion", "-ily", "pre-" &C) may be cued simply by the presence of the letters in the word. This is similar to, but distinct from, word-frequency effects.
Certain combinations of letters may increase the perceptual effort required because they have strong internal coherence, but happen to be misleading in the particular case.
Before pronouncing on the likely differences between languages, I think it would be more worthwhile to obtain a clear and thorough understanding of the roles of the above factors, and similar ones, in the process of sentential scrambled word-perception.