> People commonly theorize that virus came from rump parasitic cells. I have an alternate theory, that I've never seen articulated elsewhere: Viruses might be inter-microbial biochemical warfare gone wild; Among primitive and not-particularly robust cells, one competitive strategy might be to package toxins or other disruptive molecules into capsids that cleverly get past the 'enemy' cell membrane, perhaps aimed simply at disrupting the initially brittle reproductive process. Packaging random genes in with these capsids would be the beginning of the Horizontal Gene Transfer function that viruses perform in the ecosystem. A refinement would be to add in the specific genes for producing capsids as a deliverable so a subverted 'enemy' then takes out several neighbors (but 2nd gen capsids don't have to necessarily contain the same genes in order to accomplish their 'mission', or even be entirely identical to the 1st gen, as long as they work). A further refinement would be increasing fidelity, and packaging the responsible genes into the 2nd gen capsids produced by subverted cells so the 'weapon' can spread exponentially.
And voila, the 'weapon' has now escaped the originating organism and can evolve independently parasitizing the target. Each incremental step can at least hypothetically provide an evolutionary advantage so the whole thing doesn't have to spring forth fully formed.
Depending on how widespread variants of the original strategy were, viruses might have independently emerged more than once.
If I'm right, we ought to still be able to find cells that attack or suppress competitors with capsid-like molecules that don't particularly resemble existing virus lineages, possibly containing non-genetic payloads of various sorts.
We might also find bacteriophage endoviruses that as their 2nd stage only produce capsids sans-genes or that otherwise aren't identical to the endovirus, or ones that aren't very good at packaging up the correct genes, or that are much better at disrupting cell reproduction than hijacking it, or that the subsequent copies are only low-fidelity ones that can similarly penetrate the host but not spread any further; basically any indicators that the mechanisms for penetration of the membrane, attacking the cell, and even for making poor 2nd gen copies, have been evolving longer than capabilities only needed for indefinitely repeated copying, retransmission, and reinfection.
I suspect though that the oldest reconstructible endoviruses will no longer be capable of penetrating current members of their host cohort, which tells us little about the virus itself, but may offer clues about whatever ancestral cells were still prone to infection were like.