Two accounts on the same plan behaving differently?

A user reports that two accounts on the same "pro" plan, of similar age, are behaving differently. One account consistently uses "thinking blocks" for every response, which improves information tracking and reasoning for projects. The other account only uses thinking blocks for the first two replies, leading to poorer project reasoning and tracking, despite identical settings and explicit requests for thinking block usage. The user is seeking an explanation for this discrepancy.

This report highlights an inconsistency in AI model behavior on the same plan, unlike typical service expectations where identical settings yield identical results.