Pain only makes sense if an animal can learn. It is a teaching mechanism to modify future behavior. A very simple creature may act purely automatically, not even instinct but raw dna-programmed responses. Pain makes little sense there as there is no learning to be done. I doubt bacteria feel pain.
It sounds weaselish but it depends on what you count as pain. If you count 'response to injury stimulus' then it just needs to be a productive action in response to the injury. Plants feel pain by that definition. Bizarrely plants can be anesthetized which freezes autonomic responses like rotating to face sunlight or opening of flowers. For an example of a useful injury response without learning: something like 'invest energy in upping the poison production if something bites my leaves' would be a useful response with no learning involved.
Of course under that definition you could call alarms that go off when a window is busted a form of pain for the building. Which I suppose highlights just how loose our definitions really are. I'm no pain semantics expert but I expect rightfully or wrongly our definitions of pain are animal neurology centered.
Evidence suggests bacteria can learn. I'm not suggesting you should feel bad about taking an antibiotic but cellular life is more complex than many people imagine.
Wouldn't modifying current behavior be sufficient to justify the existence of pain? The mechanism can be a useful adaptation even if a single organism cannot learn. Maybe the default state of bacteria is pain and they can only relieve it temporarily. The ones that survive to multiply are the ones best adapted to relieving pain.
The entire purpose of experiencing externally induced pain is to cause a reaction before the threshold of permanent damage is reached. Given our imprecise biological systems, that threshold needs to be crossed with a safe margin of error. It stands to reason that reactions to avoid threats to the the continuation of life are valid, regardless of the presence of nociceptors.
Internal pain (headaches, etc.) may require some manner of cognition to associate behaviors with the result, assuming they have a cause other than flawed biology.
