This is a type of mutation which occurs with DNA during recombination. From a pair of equivalent ("homologous"), but slightly different genes, a pair of identical genes results. In this case there is a reciprocal exchange of genetic code. 

In genetic genealogy, the term is used particularly concerning similar seeming events in Y chromosome DNA. This type of mutation happens within one chromosome, and does not involve a reciprocal transfer. Rather, one homologous segment "writes over" the other. 
The mechanism is presumed to be different from RecLOH events in autosomal chromosomes and probably like gene conversion, as it has 
been witnessed in bread mold. 

During the mutation one of these copies overwrites the other. Thus the differences between the two are lost. Because differences are lost, heterozygosity is lost. 

Recombination on the Y chromosome does not only take place during meiosis, but virtually at every mitosis when the Y chromosome condenses. Recombination frequency even exceeds the frame shift mutation frequency (slipped strand mis-pairing) of (average fast) 

Y-STRs, however many recombination products may lead to infertile germ cells and "daughter out". 
* * * end of wikipedia quote * * * 

continuing, from John McEwan - "When Y chromosome STR databases are searched for twin alleles at 3 or more duplicated markers on the same palindrome (hairpin), e.g., DYS459, DYS464 and DYS724 (CDY) on palindrome P1, then a higher than expected proportion of 9-9, 15-15-17-17, 36-36 combinations and similar twin allelic patterns will be found. PCR typing technologies and new markers have been developed (e.g. DYS464X and DYF399) that are able to distinguish that these results are a combination of gene conversion or RecLOH events and deletions caused by unequal recombination. The true rates of both processes are not known, but for the P1/P2 palindrome where deletions and gene conversions can usually now be distinguished the deletions are more prevalent. The frequency of these events has not been well estimated, but appears to be intermediate between the mutation rate of STRs (~1E-4) and SNPs (~1E-8). However, the outcome is the same: what appear to be 3 separate, often multi-step mutations in different markers is in fact one single event. So a 9-10, 15-16-17-17, 36-38 haplotype can change in one recombination event to the one mentioned above, because all three markers (DYS459, DYS464 and DYS724) are affected by one and the same mutation. Appropriate allowance for this must be made when comparing individuals at these markers. Genetic distance calculations commonly used for single copy markers may not be appropriate." 

The ISOGG glossary explains recLOH as "When a section of DNA on a marker is missing, that marker is sometimes repaired by another marker filling in the missing DNA with its own material. This is referred to as a "recLOH event" and is usually observed with multi-copy markers like 385a and 385b, and is also common in the 464 set. The recLOH event causes the allele values to match 11-11 instead of the more common, 11-14 that you see in R1b." 

Here's Thomas Krahn's explanation. http://www.dna-fingerprint.com/modules.php?op=modload&name=Sections&file=index&req=viewarticle&artid=9