BLAST searches online are limited to the total amount of CPU time per individual search. This is to ensure that any single search cannot adversely affect the service. The CPU limit can be reached by searching with very large sequences, a large number of smaller sequences, or searching with sequences that have low complexity or repeat regions when the default filtering is turned off. This can be exacerbated by choosing translated BLAST searches like tblastn or tblastx. To eliminate the error you can try searching with a smaller number of sequences, or limiting your searches to one or a small set of organisms. You can also increase the word size of the search. Finally, large projects should use the downloaded databases and the BLAST+ binaries or run BLAST on a Cloud Provider.