My original proposition was that a shaman could 'connect' with the subtle magical world (without spell-use) through the hallucinogenic effect of mushrooms:
". . . from hunted mushrooms, [the druid can] produce a powerful, chewable ball that will grant both hallucinogenic effects and the ability to see into the ethereal plain and speak with intelligent creatures there. The practice requires the employment of a drum, rattle, gong or pipe, as well as a brief period of dancing and singing (10 rounds) following a hour of meditation. Once accomplished, the druid is able to speak directly to a companion that has recently died but remains within the time span of being raised from the dead (a period of no more than 29 days). The shaman may also ensure fertility in a woman for a period of one day or relieve a curse for the space of one week (without permanent effects) . . . Shamanism may not be practiced more than once per week and will drain 10% of the druid's maximum hit points."
I think in principal this still works, hallucinogenic included. I think it might be expanded to include a wide range of hallucinogenic effects, including starvation, isolation, substances other than mushrooms, near-death experiences and anything else that might disrupt the traditional working of the mind. Actual power gained would be limited in a number of ways. Any effect that resembled a spell or cantrip would, by definition, be impossible through shamanism. And I would like a precedent from some existing cultural tradition (the 'powers' I describe above - pregnancy and curses - have holistic traditions) associated with practiced shamanism. I think it is all bunk, of course - but it is an important part of human development and if we are going to accept the existence of magic, I can make room for shamans in my world actually being able to do the things they promise.
On the fungi page linked, I defined shamanism as an expert-level sage ability. That's higher than any of the other abilities I associated with tech-5. I think this means that most clans would not possess a shaman - we might suppose that such an individual would be present at the tribal level (several clans together make a 'tribe', as clans get together at specific places and times to share knowledge, materials and mating practices). It still means that the shaman would have to be part of one of those clans when they weren't actually meeting as a tribe.
We could define the chance of a shaman as 40 minus a d8 - and if the result is equal to or less than the number in the clan, it defines the clan as large enough to protect the shaman. From there, we might propose a 1 in 2 chance that this clan, and not another large one in the tribe, actually has the shaman this year.
I feel once again I need to make a point about the practicality of defining a clan of persons to this degree. After all, the party is just going to kill them all, nyet? Yet I wonder if the party would be able to recognize that, after killing a random hobgoblin in the bush, that they would realized the individual was only collecting mushrooms for their clan. If that hobgoblin turned up to have a flint and steel, would the players wonder how the remaining hobgoblins will light a fire that night? Will the players look at the age of the hobgoblin and make a connection about this being the clan's best forager? What if they find a drum - will they realize this poor dead fellow is also the clan's entertainer? Will they feel bad about killing him?
I doubt it - particularly thinking of the semi-munchkins among my players. They would probably enjoy the notion that the poor hobgoblin children will go to bed that night without the soft rattle of Uncle Ook's drum. "Be good for them," they'd say. "Toughen them kids up."
Nevetheless, I feel I'm on the edge of some discovery here - so we'll keep at it. Even if it doesn't pan out today, I may have some epiphany five years from now and remember when I went through this experiment.