One of the most unusual and striking examples of an active TE benefitting its host genome involves the telomeric transposons of Drosophila. All Drosophila species lack telomerase and instead have telomeres composed of head-to-tail arrays of specialized retrotransposons. These TEs are from the Jockey clade of LINE retrotransposons and were first characterized in D. melanogaster, where 3 telomeric TE families were identified: Het-A, TAHRE, and TART. We previously showed that TART captured a portion of the host piRNA pathway gene nxf2. Antisense piRNAs are produced from the nxf2-like region of TART, which target their cognate host gene for silencing. This strategy represents a novel form of counter-silencing by TEs. Here, we have surveyed long-read genome assemblies of over 100 species of Drosophila, identifying a total of 372 telomeric TE families, including two completely novel clades. Surprisingly, we find that capture of piRNA pathway gene fragments has occurred independently at least 8 times across the genus, suggesting that this counter-silencing strategy is relatively common among these TEs. More generally, we show that these telomere specialized elements evolve rapidly and dynamically. For example, we find that the telomeric TE gene tree is highly discordant with the host species tree, which is best explained by a large number of horizontal transfer events, as well as TE family extinction. In fact, we find that telomeric TEs have been lost completely at least 10 times across the Drosophila phylogeny. We additionally discover a novel ORF, which has replaced the POL gene in a subset of these elements, as well as an instance of convergent evolution of telomere specialization. These results provide unprecedented detail into the evolution of these unusual TEs and highlight several novel mechanisms by which they evolve in conflict with their host genome despite the essential telomere function they provide.