The problem:
We send a status call before each call to t5memory. This was done because memory can be in inappropriate status for processing the query. But t5memory now replies with its status in case it is busy and can not handle the query.

Solution:
Do not query for status before each request to t5memory (fuzzy,concordance etc.), but instead check for error containing current status in responce. This doe not apply to TMMaintenance, there we can keep calling t5memory as is.