You could do something like this, a bit fiddly, but generally functional.
| makeresults
| fields - _time
| eval data="fieldnameA|fieldnameB|fieldnameC
|valueA1|valueB1|valueC1|
|valueA2|ValueB2|valueC2|
|valueA3|Valueb3|valueC3|"
| makemv tokenizer="([^\n]+)\n?" data
| mvexpand data
| eval data=trim(data, "|")
| rex field=data "(?<f_0>[^|]*)\|(?<f_1>[^|]*)\|(?<f_2>.*)"
| streamstats c
| eval d=split(data,"|")
| foreach 0 1 2 [ eval val=mvindex(d,<<FIELD>>), tmp_<<FIELD>>_{val}=if(c=1,val,null()) ]
| foreach tmp_*_* [ eval <<MATCHSEG2>>=f_<<MATCHSEG1>> ]
| where c>1
| fields - d data val f_* tmp_* c
Up to the rex statement is setting up your example body, then
- the rex will extract 3 fields - if this needs to be more flexible, it will need some more work
- streamstats up to first foreach is then moving the header fields to their own temp field names
- second foreach is then copying the values to the correct field names
- then cleanup original header and working fields
There are possibly other ways to achieve this, but this one seems to do what I believe you are wanting to do.
If you need to increase the field count, then change the rex statement to create f_X field names and add in the extra field numbers in the first foreach statement 0 1 2 ...
