I'd like to see more isolating conlangs with some polysyllabic roots instead of tones.
With an average number of consonants (20) and an average number of vowels (5), a language could have 10,000 CVCV roots or 1,000,000 CVCVCV roots or 4,000,000 CVCCVC roots.
If roots could take any of the forms CVCVC CCVCV VCCVC CVCCV VCVCC, there could be 5*20,000 = 1,000,000 of them.
There are natlangs whose roots almost all are CVCV in form; many Polynesian languages, for instance.
Edit: There are other natlangs whose roots almost all are CVCVC in form; at least one Native North American language, for instance. If such a language were to have average numbers of vowels and consonants, it could have around 200,000 roots.
OTOH you're asking for polysyllabic roots
in an isolating
In an isolating analytic language every word has only one morpheme. So you might as well say that the word's only morpheme is its root morphome.
Maybe you'd want to call the verbs and nouns and adjectives, and maybe also the adverbs, "roots", and let them be polysyllabic. "Polysyllabic" may mean up to three syllables; or it may mean only up to two syllables.
Conjunctions, and probably adpositions, and likely pronouns, and maybe also adverbs, might be "particles" instead. Maybe you'd want most of them to be shorter.
There'd be many minor parts-of-speech. There'd probably be a hell of a lot of auxiliaries, for instance; these would also be "particles". There'd also be articles and quantifiers and determiners and degree-words and so on and so forth.
I don't know whether you'd allow apophony or ablaut to operate on a "root", or whether that would be a violation of isolatingness or of analyticity, or whether instead you would count that in with a "tone" system as something you'd rather avoid.