Malaza (P)

4 Sep 2026: Regiment leftover "SA Army" cleared -- that's the service branch (Army, as distinct from Navy/Air Force), not a specific regiment; the actual unit ("4 South African Infantry Battalion") is already captured in Unit. Per John's decision, the 52 records with no better Unit info were left as-is.
16 Sep 2026, ServiceNo contamination sweep: the stored ServiceNo '94971892' was CLEARED. It was not a reading of this man's record. PersonNumbers 31534 to 31535 -- 2 consecutive records -- all carried the identical number, which belongs to the record at the head of that run (31534, Malatji, Vusimusi Stephen (V S)). A value repeated down an unbroken block of PersonNumbers is a forward-fill import artefact, not men who shared a service number; the pattern holds across 850 such runs on this roll and is independently confirmed by Mike Sandham's research log, which attaches such numbers to the head of the run 167 times against 6 contradictions (all examined). No replacement is guessed: this man's real service number is simply not known, and a blank field says so honestly where another man's number did not.
25 Sep 2026, Unit2 service-branch sweep: the stored Secondary Unit "Army" was CLEARED as a duplicate. It was not a sub-unit, it was a service-branch word repeating what the Branch column already says. Secondary Unit is for sub-units -- a company, a squadron, a commando -- and this record's branch of service is unchanged at "Army".
Database-wide FirstName/Age/Unit contamination sweep, 28 Sep 2026 (John's hypothesis, tested and confirmed on 3 sample groups before running at scale): FirstName 'Vusimusi Stephen', Age 19 and Unit '4 South African Infantry Battalion' were forward-filled from PersonNumber 31534 down a contiguous run of 2 records -- this record's own LastName ('Malaza'), Initials ('P'), ServiceNo ('') and DateDeath ('1992-01-01') are all genuinely distinct from every other record in the run, confirming a different real person whose FirstName/Age/Unit were simply never captured (source pattern consistent with a SAPS memorial table using rowspan for Name/Age/Province, mis-copied by the scraper onto every row). FirstName, Age and Unit CLEARED; no replacement guessed. CauseDeath NOT touched -- it follows its own, narrower sub-block boundaries within this run and carries real distinguishing information.
CORRECTION, 28 Sep 2026: the earlier note on this record wrongly described Unit as forward-filled/contaminated alongside FirstName and Age. On review, Unit values proved to be common, genuinely correct data shared by many unrelated real people throughout this dataset (e.g. 'Western Cape' alone appears on 195 scattered, unrelated records with 111 distinct surnames) -- Unit was NOT part of this contamination and has been RESTORED from the pre-sweep backup. FirstName and Age remain correctly cleared -- both were 100% correlated across every contaminated run and independently confirmed absent on SAWGP's own record for spot-checked samples.