As a graduate student in linguistics, I had the opportunity to learn about creole languages. By around the year 2000, I had come to the conclusion that a world language with a truly international lexicon and a simpler phonology and grammar than that of Esperanto was actually possible. Many of the features in Globasa are based on or inspired by the most typical features found in creoles, such as their analytic grammar.
By the time I encountered Neo Patwa around 2009, and Pandunia in 2017, I had already developed the following ideas for a hypothetical full-fledged worldlang that didn't sacrifice expressiveness and clarity, in other words, a worldlang that could actually be a better alternative than Esperanto at all levels:
Truly international lexicon
Use of cross-linguistic onomatopoeia (source for all of Globasa's onomatopoeic words)
Choosing easily distinguishable function words and avoiding minimal pairs
When I encountered it in July 2017, Pandunia seemed like the closest worldlang to what I was envisioning. The phonology, however, seemed a bit more complex than it was necessary (particularly the use of voiced and voiceless stops in word-final position) and there were other issues that I felt needed to be addressed, including similar-sounding words, such as the pronouns. After lengthy Facebook discussions with Risto Kupsala (Pandunia's creator) regarding its phonology, Pandunia's phonemic inventory was changed back to a simpler, earlier inventory. Afterwards, upon closer inspection, I realized that Pandunia lacked the necessary morphology/syntax to disambiguate simple sentences, let alone complex sentences. It was around this time, only a few weeks after encountering Pandunia, that I decided to part ways and begin my own project.
As I began developing the project, I realized that the best way to start was with the most difficult part: function words, including grammatical and derivational affixes. In doing so, the first question I had to address was what to do about content words, specifically, whether to go with classless words while relying on syntactic particles (which I assumed would be Pandunia's approach) or whether to assign word classes and probably use affixes to switch between them. I knew I most definitely did not want to introduce vowel endings, such as in Esperanto. This would've represented a clear deviation from the typical creole language model. I opted for the middle ground between the two extremes of the classless word approach and the vowel-ending approach: adopting two major word classes, noun/verb words and adj/adv words. The reasons for rejecting the classless word approach are complex and have been discussed with Risto elsewhere.
I initially anticipated spending about one year to come up with the entire core grammar, at least 500 total word roots and affixes plus a simple website. Instead, it took about that long to decide on the form for all function words and affixes, while developing the framework for the core grammar. It took me another year to complete the core grammar, to come up with a lexicon of 1,000 roots and to complete the first website, all while continuing to make adjustments and changes. Before publishing, I wanted to be sure the project was at least 95% stable.
The name of the language was originally Geokreol. Later on, I considered Gayabas. Initially, all truncations omitted the final vowel of the root word (basa → bas, for example), but once I decided that truncations need not be systematic, I came up with -sa for naturalistic language names (Espanisa, Fransesa, Turkisa, etc.), which led to the merged root Globasa. I decided to publish Globasa on July 26th, 2019 as a nod to Esperanto, the date in 1887 when the Unua Libro was published. After its publication, we ended up making more changes and adjustments than I had initially anticipated, so it's possible that Globasa was under 95% stable when first published, but certainly not under 90%. See Globasa's Stability.