After a few thoughts about float paths not making much sense I would like to add other comments about the analysis of multiple paths.
http://www.cassellconsulting.com/consulting/multiple-float-paths-in-p6.html
http://ronwinterconsulting.com/Longest_Path_Value.pdf
In theory simple computation but short if missing the statistical element.
I would not trust an algorithm that order multiple paths based on an arbitrary measure that does not take into account the probabilities of each path be delayed and by how much, too much is missing.
Google for Criticality Index and you will find even this metrics is limited and alone can be misleading.
Here the scientific approach would be using Statistical Modeling within your CPM software, it can be via Monte Carlo or or other Statistical modeling. Beware that the application of Monte Carlo using external software might invalid the analysis. This will provide you with the tools to measure the probabilities of success as you change your strategies.
I would like to highlight Ron Winter argument about how out of sequence activities makes it difficult to trace the critical path, a very interesting article you should read. If you software is capable of filtering for active driving links then the identification of completed out of sequence activities in the critical path shall be no problem. Do your homework and remove all out of sequence occurrences, even on activities completed, this will eliminate the need for esoteric procedures. In any case it is not always by working on the critical path that you increase your success probabilities.
Float is not real, useful metrics but in no way absolute.