Even though I had made a broad guesstimation, I don’t seem to have been too off after checking some data.
Before replying to you I went to check some census stats to see where I was standing.
My initial assumption was considering only the working-age demographics, and that there would be synchronous spending peaks based predictably by schedule (ie. at noon, lunch).
The best case to study would be India where cash is still king, even after three years of the controversial forced demonetization to implement a cashless economy.
Just considering the working age demographics in India:
Ages 15-64: 741,248,288
Ages 64-over: 61,955,950 (including this group as a group that would be resistant to cashless transactions)
Total: 803,204,238
If we assume that all of this people does at least one transaction at lunch, that is the minimum amount of transactions that would be happening more or less at the same time.
But this is only one country, at the same time zone (UTC+5:30) we have Sri Lanka:
Ages 15-65: 13,625,110
Ages 65+ : 1,602,663
Total: 15,227,773
India and Sri Lanka are in a weird time zone, it is 15 minutes ahead of Nepal (UTC+5:45), and 30 minutes from China (Xinjiang Time), Russia (Omsk Time), Kazakhstan, Kyrgyzstan, Bhutan and Bangladesh (all of them UTC+6)
Considering how close they are, it would be fair to also consider that lunch time would greatly overlap between these countries, so let’s add these people as well:
(Working age + elders)
Nepal: 16,885,164
China (Xinjiang and Tibet): 20,250,213
Russia (Omsk): 1,113,241
Kazakhstan: 12,366,257
Kyrgyzstan: 3,826,342
Bhutan: 503,394
Bangladesh: 102,905,000
That gives us a grand total of 976,281,622 people under relatively the same time zone.
Moreover, to be complete we would have to also add countries -/+ 6 hours to include those that would be having breakfast and dinner, which would be happening at the same time that these people are having lunch…
So we would be getting somewhere in the ballpark.
A conservative assumption of buying one snack at meal time would then be equivalent to about 1.5 billion transaction within half an hour to an hour, and who knows how many hundreds of millions would be overlapping at the exact same instant.