Sometimes hitting limits means a new approach.
In my last post I was very disturbed by the attitude that hitting limits (or wanting to know about the box you are in) means you are doing something "wrong". Yet, the main examples used actually do represent reasonable limits... 50 nested windows, 25 level menus and similar are most likely pathological code. In fact, such limits can help avoid denial of service attacks that consume system resources via recursive calls. Nevertheless, even limits like these leave me uneasy. Unexpected uses do not necessarily mean abuses of a system. I believe any hard coded limits are suspect until proven otherwise. When you can limit things based on dynamically allocated resources, trust levels and system usage patterns, hard limits just look arbitrary.
Still, the circumstance that Jeff describes does seem to be pushing a limit that few would encounter.
If you have 39 IE7 windows, something probably has spiraled out of control (like malware). 47 tabs across all instances on the other hand does seem a strange limit. Jeff does point out that Firefox allowed over 100 tabs, although it then became unstable. (Update, apparently this limit is imposed by the Desktop Heap, which harks back to the old limits of the System Resources of old. Here is the way to expand the heap courtesy of Ed Bott).
I suggest that a different methodology be used here: Jeff seems to open every relevant topic and leave it open while doing research. This seems a brute force approach to a problem that could use some help organizing these tabs.
The simplest solution is to bookmark to a folder interesting topics. However, this is a limited technique is that annotation of such bookmarks is weak. Social bookmarking sites such as Delicious work fairly well as they allow tagging and navigation through bookmarks. The issue here is that the volume of links becomes overwhelming to scan through and tagging is challenging to make granular enough to represent multiple clusters that share some, but not all links in various tag groups.
Instead, I usually use a product that can handle bookmarks and my brainstorming at the same time such as The Brain or Mind Manager. Both do this task very well; I'm sure other mind mappers can be substituted here.
If you haven't used mind mapping software before, it does take some time to get used to. The simplest approach is to treat them as an outline that happens to be 2D in layout. However, products like The Brain and Mind Manager have the ability to interlink in more complex ways and focus on parts of the map in a way that makes "random" collections of data coalesce into meaningful clusters naturally. In particular, when just searching around on random topics, The Brain has a "recently visited" view that makes these kinds of research projects even easier (as your recent entries are probably what you are working on now) and a strong search capability for hunting through large maps. I find it more useful for undirected research. Mind Manager on the other hand moves easily from brainstorming to project estimation, making it useful for collecting facts about upcoming projects.
So, although I decry the celebration of arbitrary limits, in this case I think it is a matter of methodology. When researching for a new topic, I often find myself using tools like these to quickly organize a disorganized pile of links into clusters of related topics. By storing links in a product such as these (or even simply making an outline in a word processor with the links in it) the need to maintain a large number of open tabs vanishes... and the future reference to the material is vastly simplified. As a side benefit, I find my focus is sharper when I'm presented with a cleanly organized set of topics rather than a wall of tabs.
Still, the advice in the linked article about increasing the Desktop Heap does seem useful in these days of 2-3GB 32 bit desktops... why suffer errors that hark back to the dark ages of 16 bit resource heaps.