Thank you for the excellent, detailed questions regarding geography for the census samples of Vietnam.
First, only the 2009 sample contains the district codes because the sample percentage for the long form is 15%, while earlier samples are 5% (1989) and 3% (1999) and thus are not designed to be representative at the district level. At this time there is no way to obtain district identifiers for 1989 or 1999 samples…
Second, the household identifier is an arbitrary serial number, and thus is not suitable for tracking households across censuses. To protect statistical confidentiality, there is no way to track households from one census to another in any IPUMS-International samples.
Third, yes, mgrate5, in the case of Vietnam, refers to province and district as the major and minor units, respectively. This is documented on the comparability tab of the variable: https://international.ipums.org/international-action/variables/MGRATE5/#comparability_section
Let me add that, later this year, we intend to request short form samples to facilitate analysis at the district level. The General Statisiics Office of Vietnam is one of the founding members of IPUMS-International, so hopefully the negotiations will be fruitful.