The first habit is efficiency, the avoidance of seemingly redundant things, such as multiple forms of agreement. No human language is perfectly, or even partially, efficient. Indeed, redundancy in a spoken language is a positive benefit, since language happens in a noisy medium. Nonetheless, nearly all my languages avoid certain features. For example, I typically use either strict head marking or strict dependent marking. Rarely do I allow overlap, which is somewhat unusual.
At this point Bixwá is not going to chage in this regard, but I decided in Tsrai that I will mark the plurality of the subject in both the subject and the verb.
Gad kóis A man sleeps
Órói këskóis men sleep (reduplication for the verb, and the plural of man happens to be irregular, órói)
This double-marking of plurality is quite redundant — but also quite common in human languages.
The second habit is the tendency to avoid ambiguity. In beginning (or engineered) conlangs, this commonly presents itself in the form of perfectly identifiable word classes. Every derived adverb in Esperanto, for example, ends in the same letter, -e. In Láadan all the speech act morphemes begin in the letter b-. In one of my earlier languages, Mavod, all the tense markers begin with the sound sh-, and other word classes exhibit the same patterning. I've been getting away from this habit, which I now find rather dull, and in Bixwá I have evidentials I would once have felt slightly uncomfortable about and avoided. They are suffixed to the clause, usually landing on the verb:
|-aazh||obvious, "of course, as we know"|
|-jíín||report from untrusted source|
Only the two report evidentials show any family relationship. Further, there are plenty of verb stems that end in -sh and -xw, a situation I would once have avoided.
The third habit I call "orthogonality," which presents itself in multiple guises. I'm abusing the mathematical concept of orthogonality, but by this I mean the tendency to ensure maximal distinctiveness in whatever feature we happen to be working on at the moment. For example, we might mark all tenses using the same morphological pattern, or we might fill out all the possibilities of tense and aspect if we've decided a language needed that. Some natural languages do work this way, but plenty do not. Navajo, for example, is most preoccupied with marking aspect rather than tense. Nonetheless, it does have a future. For Tsrai, I've decided the verb has only past and non-past tense marking.
In Bixwá I have deliberately mixed aspectual marking among different sorts of morphology. For examples, the aspect "begin to" is marked with the aspect prefix koo-, but if you want to talk about a stative verb (what English uses adjectives for) entering a state, one uses a mix of the perfective prefix ho- and the outward direction prefix, cho-. For example, jed is be cold. To say "it got cold," you use chohojed (cho-ho-jed out.away-PRF-be.cold).
I have also mixed up certain matters of syntax and pragmatics in different places in the language. For example, Bixwá does have some conjunctions that work as in English, but it also has some preverbs that do jobs much like conjunctions (preverbs are mostly adverbial particles that must occur in a fixed relationship to the verb, and which don't participate in other morphology). For example, the preverb wil indicates continuation, "then, and then, next,"
Né wil kora-n ho-deezh.
1SG PREVERB:then dog-ACC PFV-see.
And then I saw the dog.
I have also grabbed the preverb sa', thus, there, at that time, for a discourse function — it asserts narrative integrity, that the statement is related to the current discourse. This is useful when something surprising or unexpected is being said, but I find I use it a lot in any narrative. I've also grabbed the preverb jééx, which carries "emerging, up and out" ideas, to be yet another way to indicate an inceptive or inchoative sense for transitive verbs (distinct from the stative verb cho-ho- pairing).
Some years ago Jesse Bangs posted to the conlang mailing list The Conlanger's Rant, which suggested among other things conlanging schools and conlang criticism, himself being a devotee of the "naturalistic" school. Along with a lot of conlangers, I was rather turned off by this idea. Among other things, I've never aimed at perfect naturalism in a language. I certainly had no desire to belong to a Naturalistic School of Language Creation, and still do not. But my tastes in how a language I'll find interesting ought to be constructed have changed, with the surprising result that I've been pulled more in the naturalistic direction, not from a desire for naturalness, but from boredom.