The peer is too busy to handle the request right now (DIAMETER_TOO_BUSY) -- a load problem on that node, not a rejection of the subscriber.
Where you see it
In the Result-Code or Experimental-Result-Code AVP of a Diameter answer (S6a, S6d, Gx, Gy, Cx…).
What to check
Ask the operator of the node that answered to check its load; the request can be retried to an alternate peer.
Look at the Origin-Host of the answer: an intermediate agent (DRA/IPX) and the home HSS mean different owners of the problem.
Got a trace with Diameter 3004 DIAMETER_TOO_BUSY? Drop it in the analyzer: it finds the step that really failed first, which node sent the code and what to ask the other side. Captures, test-system PDFs and vendor exports. 5 free analyses a month, no sign-up.