It sounded like s/he was interested in asking scientific questions, not assisting others to do so. If that's the case, there's just no way around needing control of a funding stream.
You can control by influence. I know of examples where the "research scientist"/"research assistant professor" was never a PI on any grants, but is able to convince the study PI to do research in a certain way, and eventually be main authors on key papers. That can be a path forward, but you are right highly unusual. Salary support is an administrative hurdle that's somewhat orthogonal to the control factor.
My hope of even making such a thing possible would be through developing really deep data analysis skills and playing w/ publicly available or otherwise acquired data sets. There is a world of "Neural Data Science" emerging which seems like a much better means of staying "close to the science" without the tedium of actual bench work, but a good deal of excitement and thinking about things. Also if you know how to analyze big Neuroimaging, calcium imaging, or spiking data sets, seems like you'd be useful to many groups in a flexible way where you could be a clinical to cover yourself and do real 'science' on the side. Given I'm really interested in Psych, this is something I'm keeping in mind. but echoing what
@tr and
@sluox have said, it seems neither desirable nor feasible to just be an research tech for the rest of your scientific career.
It's not easy to do even for software. There are R01s for secondary analyses of large public datasets, and PIs who occupy those grants will be recognized as experts to do this work. It's also more plausible if you develop a relationship with the consortium that acquires this data in the first place--it'll be difficult for you to even understand the data without input from study investigators. You also don't need to do this work for free. There have always been many active career and project, and even massive center grants and cooperative agreements for methodological development, modeling, etc.. It'll be difficult for you to develop credibility for "data analysis skills" as a "gentleman scientist"--you'll do a lot of work, some of which may even be highly valuable, but it'll not be appropriately recognized. It's also common to "step on people's toes", duplicate effort without coordination. This is why people just constantly judge you based on your biosketch. On the other hand, when you have a relationship, money just pops up without even an explicit application via things like administrative supplements and budget revisions. Do you really think it's hard for senior people to find money for your 50% FTE when the trial costs 50-100M? What is credibility but a bunch of people know who you are and what you do? But in order to develop relationships, you'll soon find that the easiest way is through existing institutional pathways, get a fellowship, write some papers, write a K, declare yourself as an expert in data science of X neural data, then apply for bigger grants, etc. etc. This is what "the system is gonna system" means. If you find this process intolerable, it won't be solved just because you transitioned your science from wet to dry, from animals to humans, or any of that.
Technical skills are a commodity because the training can be scaled. Entry-level data analysis skills are cheaper now, given the millions of coding schools and bootcamps that are popping up, and you are competing with a large pool of new PhD grads in an actual quantitative field, if you don't have that. If you are talking about entry-level technical skills, MD skills are [again] more valuable. Depth, which typically considers both soft and hard skills, cannot be scaled and is what's actually valuable.
It's the same in any industry. In the 80s maybe you can be a high school drop out and sell your software to Microsoft for millions, but these days it's more reliable to be a L7 Amazon engineer for 500k a year. What do you think an L7 at Amazon does? Not writing Python code--it's meetings to write documents similar to an R01 application: product development plan, timelines, aims, milestones, deliverables, budget, blah blah and then MANAGE their coders to actually write the code, once it's shopped by a committee of people in charge of the money. Many such proposals get turned down. What do you think a director of clinical development at a big pharma does? Same thing. Generating documents to be read by people who read documents and make a decision about money, and then do more writing to formalize the process where the CRO does the "on the ground" work for you. LOL I don't know why people are so against being middle management. I think I'm just getting old...