One of the main attractions of Logseq is the fact that your data is stored in plaintext files, ensuring longevity and a certain degree of portability for your data. One thing that breaks this are the random numbers used in block references. I suggest that instead of (or in addition to) these random number, block references also use a string of letters from the original block text. This will allow users to search for the block even if the block reference itself breaks for some reason. Many have already seen block references break due to updates or sync issues, so this will allow users to re-establish those blocks. It could be combined with a feature to “re-link” a block ref by searching for the text and confirming that it is the proper block.