Much depends on the setting and the GM. You're right; logic dictates that you have access to resources that will support the skill improvement, and so, if the GM is allowing IPs every session then it makes sense to limit skill picks to those that can be supported.
However in many cases GMs award IPs at the end of the story, which then assumes that characters are in downtime and able to find access to things they want/need.
The IP award system is structured in the way it is for several reasons.
1. To allow characters to develop holistically. If you limit IP spend to skills recently used, then you'll find similar skills being improved at a much faster rate than others. What do you do if your character didn't use a skill at all during an adventure (it can happen)?
2. To get away from the old BRP trope of only increasing skills successfully used. Again, what if you failed in a skill roll due to simple bad luck? In a session where combat was a principal focus, only combat skills would get the chance for improvement.
3. To encourage players to improve skills reflecting their culture and profession, rather than just those that seem obvious for direct game use.
Also many skills don't need additional resources: just time for practice and consideration.
Of course, it can work as has been suggested, but the intent of the rules - and this is why there's no specific statement that only recently used skills can be improved - is to create flexibility and diversity. Is it too generous? No, I don't think so. Much depends on frequency of awards rather than the way in which skills are improved, or even which skills are improved. In my recent Arthurian campaign, which operated on highly compressed timelines and IP awards every two sessions, it took some time for characters to reach dizzying skill heights, and the players quickly realised that diversity of IP spend was very important.