Dead lock by using IsolationLevel.RepeatableRead

This comment was long and may answer your question:

IsolationLevel.RepeatableRead is very restrictive. It ensures that any data is locked at the point of being read. Any select transactions you have will also lock the row until the transaction has been completed.
so (A) reads, (B) tried to read but deadlocks, (A) completes. (B) will be discarded as it holds fewer resources.

I prefer IsolationLevel.Snapshot when I have larger transactions. It's risky but it work in my case:
(A) reads, (B) reads, (A) writes, (B) is unaware that the data has changes in the mean time so (B) writes over it.

I will almost never have two transactions writing to the same rows by design. Where they can do this is not critical and I'm happy to let one overwrite the other. I will often read from rows that are being rapidly updated but because I poll my data very often I don't mind if my data is out-of-date by a second or two.

Research the isolation levels:- Consider using specific isolation levels for specific transactions:- It may be that you require this isolation level in which case add a retry process.