Dear Pierre,
This isn’t an error: there is no rule stating that you must start again at “1” for each new parameter.
I even thinks this also applies to xxSEQ in general. For example, if I have 9000 obervations for subject “SUB001” and give them a value of “1” to “9000” for xxSEQ, and then start with subject “SUB002”, it is completely valid to assign a value of “9001” (or even “19001”) to the first record of “SUB002”.
The SDTM-IG is very confusing there (like in many other places ). For example for LBSEQ it says “Sequence Number given to ensure uniqueness of subject records within a domain. May be any valid number”.
I consider “sequence” and “any valid number” as a “contradictio in terminis”.
So if you start with “18000” for the first record for subject “SUB001”, and the second record gets “17996” and the third “17998”, that would i.m.o. also perfectly OK. As long as xxSEQ is unique within the subject, you should be ok. But … reviewers will probably not like it …
Please also take into account that examples in the SDTM-IG are not anything more than … “examples”.