Consistency assumption on final pending block (per slot) between of EU and US relays

[nakedkoalacapital]

My program queries for pending next block inside aloop via eth/v1/builder/header{slot}/{parent_block_hash}/{validator_pub_key}

Should I make the assumption that the final pending block ( the last version before slot ends) hash will always match between the 2 relays ?

Based on some preliminary testing, it does seem like this is the case.

But I do get some cases where the expected consistency is not observed. ( the final version of block hash from EU does not match that from US). I am not sure if this is because my assumption is wrong or that my polling loop missed a update before the slot rolls over to the next one.

Can the team comment ?
If I am interested in polling the latest best pending block hash from ultrasound, would it be sufficient just polling 1 of the EU/US relay ?