I am sorry, but your friend is generally wrong on his point that small changes is good for search performance. This is, simply, not how search works.
Here are a few points to consider.
Templated Content v.s. Content
Google and other search engines can easily compare more than one page to determine what HTML elements are templated, where and how these elements are used such as header, sidebar, and footer, and the relative value of the templated content.
In this, Google and other search engines evaluate actual content separate from templated content. Why? Because search engine matches to, for example, sidebar elements poisoned the SERPs with poor results. It is that simple. This was a real problem at one point. For search performance, short of a few additional HTML elements, only the content is considered for search.
How content is evaluated.
There are several semantic analysis techniques used to analyze the content one of which is topical analysis. I have talked about this elsewhere, however, for completeness, I will go over it briefly again.
Google does not make keyword matches. Why? Because many terms can be used in more than one context. For example, the term dog can be an animal, the poor performance of something, a description of a persons attractiveness, etc. Humans can put terms into context instinctively, however, computers cannot. Semantic analysis is used to understand the written word. Semantics has existed since at least the 1970s and follows how people think and communicate. It is intended to accurately understand meaning.
One of the techniques is topical analysis. Each term, where it applies, is assigned a topic category or sub-category from an ontology. Where an ambiguous term is used, disambiguation techniques solve this problem. Each logical content block is evaluated in order and evaluated for importance. The entire content, each header, paragraph, sentence, and term is given a score that very accurately defines the topic from the largest possible meaning to the smallest possible meaning using an hierarchy. For example, a carburetor is a car part. A car part would be a sub-category of a car. While this is an overly simplistic example, this example illustrates how a term fits within a hierarchy that allows for narrow or broader meaning.
This and other semantic analysis scores of the various content blocks is put into graphs typically referred to as a matrix. Content blocks in relationship with each other are overlapped using matrices of matrices. For example, this means that a header topically supports the paragraph and paragraphs that immediately follow.
These scores are what is stored within the index. Google is both clear and ambiguous as to what semantic methods are used within its index. However, Google is up front with much of what is used.
Search queries are also evaluated using semantics.
Google is clear that semantic analysis is used to evaluate a search query. Shorter queries are ambiguous while longer queries that allow for more semantic understanding are far clearer and yield far better results. The same analysis is applied to the search query as is to the content. This allows for unambiguous search queries to be closely matched to semantic scoring within the index. Where people think of keyword matches, it is the topical analysis that primarily drives search query matches.
So what does a few simple changes to content do?
Almost nothing. Any minor change to content would not or would barely move the needle as far as semantic scoring is concerned. It is that simple. Google has the opportunity to compare the changes found between the cached and fetched page. Minor changes where the content does not make a significant change in the scoring are not seen as fresh. Again, simple. In cases where minor changes are made, the index is updated, however, the net gain otherwise is little to nothing. For content to be consider fresh, the content must change. Not the formatting, not minor changes, but measurable change.