I'm NOT too sure about this, BUT.......

It seems to me that if a conlanger wants to create a naturalistic
conlang, s/he should be careful not to create a phonology and/or
root structure that's complicated. If the phonology and root
structure makes it possible to create hundreds of thousands of
possible root words, it seems to me that a naturally evolving
language would have the tendency to simplify things by either;
removing phonological contrasts that would not be necessary, or
simplifying the structure of roots. Alternatively, s/he should not
create a phonology and/or root structure that is too simple for it
seems to me that natlangs would tend to complicate matters to create
contrasts that would be necessary in a language. Basically, a
language would strive to obtain/maintain an ideal number of possible
roots. If such is the case, is there a recommended maximum/minimum
to the number of possible roots such a conlang should have?

Below are some examples of natlangs that appear to reflect this:

Why has Chinese become tonal for instance? One reason could be that
since roots are predominantly monosyllabic and simple in structure,
tones would have to arise to compensate for a lack of phonemic
opposition in words (i.e., to decrease the number of homophones).
But say Chinese developed into a language with polysyllables, then
the tonal contrasts would not be necessary. In fact, Mandarin
Chinese with only four contrastive tones has quite a few
polysyllabic roots. Other Chinese dialects with more than four tones
are more monosyllabic than Mandarin. Basically, these changes
attempt to preserve the number of possible roots.

Vietnamese is now with more certainty considered a Mon-Khmer
language. Mon-Khmer languages are non-tonal and polysyllabic. How
then did Vietnamese become tonal and monosyllabic? Basically, under
the influence of China (and perhaps Tai) it become monosyllabic and
tones would have to arise to compensate for contrastive
polysyllabicity. Similar situation as I demonstrated with Chinese:
Preserving the number of possible roots.

Proto-Tai once had three tones. Thai and Lao today (depending on
dialect) has five tones and six tones respectively. These extra
tones arose after the "Great Tone Split" when tone contrasts
developed to compensate for the loss of contrastive initials.
Similar situation as demonstrated with Chinese and Vietnamese above:
Preserving the number of possible roots.

These are all (predominantly) monosyllabic Asian languages. But a
similar picture can be painted for African languages. Many of these
(if not all) are polysyllabic. But many are tonal as well. Compared
with what I have preciously mentioned about monosyllabic Asian
languages, one would think that tones would not arise in
polysyllabic languages. However, African tonal languages have
syllable structures that are even more simple than Chinese,
Vietnamese, or Tai (i.e., CV structure instead of CVVX). Now if
these African languages were to develop more complicated syllable
structure or increase the phonological inventory, I would think that
tones would disappear since they would no longer be necessary to
create contrasts between possible roots. Basically, preserving the
number of possible roots.

All the above examples are tonal. What about non-tonal languages?
Well... ever wondered why English roots are predominantly
monosyllabic? I'd reckon that its because of the extrememly complex
syllable structure. If English were to developed a much more
simplified syllable structure, then many things could happen; tones
could develop, or the language becomes more polysyllabic, or new
phonological segments would develop. Basically, placing the
complexity somewhere else to compensate and preserve number of
possible roots.

What about minimum and maximum number of roots? Of course there are
languages that have tens of thousands of possible roots, and there
are those with very few. Philippine languages for instance, with a
bisylabic lexeme structure that is predominantly CV(X)CVC, have a
very huge number of possible roots. If we are to believe
Austronesian prehistorians about the migration of Austronesians, the
Malayo-Polynesian speakers originated from the Philippines. So if we
look at the Malayo-Polynesian languages that sprung outside of the
Philippines, they appear to be simpler in root structure. Today,
Malay does not seem to allow voiced stops at the end of words, while
Polynesian languages developed an extremely impoverished
phonological inventory and predominantly CVCV root structure. In
this situation, instead of preserving the number of possible roots,
efficiency has been attempted (especially in Polynesian) by reducing
the number of possible roots to an ideal amount such that there are
JUST the amount of contrastive roots needed.

There are probably more examples that I'll recall after I send this.
In anycase, I'll stop giving the boring lecture.... 8-)

-Kristian- 8-)