Meta-comment - this is a great question. Probably there are many similar questions about difficult prioritisation decisions that EAs normally try to solve individually (and which many, myself included, won't be very deliberate and systematic about). More discussions and estimates about such decisions could be helpful.

Good question! One consideration: in many cases, mentorship may not trade off directly against direct work. Many people report that there is a limited number of hours of research/writing/'deep work'/hard thinking that they can do in a day (people often say 2-5 hours);  but they can do other, not so focussed work on top of that. This is certainly the case for me! (Not that I'm a senior researcher). I suspect this is why in academia, it's customary for professors to both research and teach - they wouldn't spend all their time researching anyway. 

So, while it's certainly possible for mentorship responsibilities to be distracting and seriously trade off against research, I suspect that with the right balance, many researchers will be able to do research at their full capacity and also do a limited amount of mentorship. 
 

This question is related to the question of how much effort effective altruism as a whole should put into movement growth relative to direct work. That question has been more discussed; e.g. see the Wiki entry and posts by Peter Hurford, Ben Todd, Owen Cotton-Barratt, and Nuño Sempere/Phil Trammell.

Like Stefan Shubert says, this is a great question!

It's brief, tightly written and high signal to noise, and has a lot of "object-level content"—like the tradeoffs between object level work and community growth, and the key value mentorship has. 

a) Has anyone ever thought about this question in detail? 

• I haven't thought about this in detail but I have a weakly held view that senior people should do more mentoring
• (without wanting to imply that I'm a "senior EA") I've thought about it / am generally inclined to think about it more carefully for me personally, I think last time I did I basically thought I'd like to do more mentoring and was bottlenecked on not having anyone to mentor (but not sure that I currently think I should do more mentoring)

b) What factors would such a decision depend on? Intuitively, the senior's ability to mentor and the urgency of the problem play a role but there is surely more. 

• For myself I might try to do a fermi, maybe something like (just for illustration, extremely rough and not thought through, etc): career capital for me (maybe made concrete with "how many hours of productive time am I willing to sacrifice for the expected career capital") + career capital for mentee (maybe "by how many days do I accelerate their career progression x how much impact will they have compared to me") - time cost for me
• I will say I think some people enjoy mentoring (and similar things) way more than others, and this probably matters a lot. Maybe in the above fermi you can apply a factor to the time cost to convert it from "actual clock time" to "counterfactual difference in time spent on other productive things" or whatever
• Maybe another factor is how disruptive it is for you to add (say) another meeting per month to your diary. E.g. if you usually have around 2 meetings per week and otherwise can focus on research, maybe adding more meetings is very costly.

c) Are there options to combine mentorship and direct work, i.e. can senior people reliably outsource simple tasks to their mentees?

• I think outsourcing simple tasks is surprisingly hard. But maybe a good version of this looks something like having an RA/PA (and maybe senior people should have more RAs/PAs).