There should be no limit on the number of kernels.
What might cause the problem is that the default name for the kernel spec is julia-1.X for each minor version, so probably one kernel spec got overridden when you installed the second 1.9 kernel.
You can manually install another kernel with installkernel (also useful if you want to have a different kernel with multithreading etc.)
using IJulia
# Let's say this is in the 1.9.0 environment
installkernel("Julia-legacy")
and then install the latest version normally with pkg> build. Here is more info
https://julialang.github.io/IJulia.jl/dev/manual/installation/#Installing-additional-Julia-kernels
You can also check if the kernels are pointing to the correct locations of the Julia executable by inspecting the paths returned from
shell> jupyter kernelspec list
In the folder corresponding to the listed kernel spec, there should be a kernel.json which contains the path to the Julia executable.
PS: IJulia (or jupyter?) seems to always append the full version number to the kernel name displayed in Jupyter (at least Jupyter lab). Not a problem, but it looks a bit silly if the name is Julia 1.9.0 1.9.0, so I chose a different name in the example 