TESTS: t7145-solver-Eckerle4.pl can finish ( sucess / Pass ) with a restart
1.) Observation:
t7145-solver-Eckerle4.pl fails ( 'fake passes' ) with:
--> -0.17215757957205108
Known failure.
in short time ( < 1 sec. ) and with few iterations ( < 50 ).
2.) different in 'playing around':
When opening the sample in a worksheet, solving, and then after 'Done' starting a new 'Solve'
2a.) a different set of values is calculated for I41:I43,
2b.) a much lower minimum of ~0.00146 is reached in I45, and
2c.) a much better 'quality' of about 9.1 is reached in K39, which qualifies for a real 'Pass'.
3.) Assume:
There is something different in calculation sequence, evaluation logic ... at the 'end' of solving than in the start?
4.) add. observation:
4a.) gnumeric-long reaches a real Pass without second run,
4b.) similar as 2.) - improving in second run - may apply in other solver sheets, is observed in 'Thurber' and 'MGH09'.