I think your "callback url" implementation is rejecting encoded nostr events which are too long ... these events can get too long (in characters) if there are a lot of relays.... the relevant part of the spec is here: https://github.com/nostr-protocol/nips/blob/3e8c33ee79e3d2eb46ca498eddadcc6ef1ef7329/57.md?plain=1#L31
Testing zaps for this note… we made six attempts to⚡zap this note, at pedromvpg@btcpay989117.lndyn.com, over a period of 19 minutes. In each case, we found that your lightning address server did not respond correctly. (The failure point was when we did a GET request to your specified callback URL: https://btcpay989117.lndyn.com/BTC/UILNURL/pay/lnaddress/pedromvpg.) Your server did not produce an invoice, thus the zap failed. We recommend that you use a cloud-based Lightning node which is optimized for Nostr zaps, to get your zaps working.
Discussion
I assume it's a bug in the nostr:npub155m2k8ml8sqn8w4dhh689vdv0t2twa8dgvkpnzfggxf4wfughjsq2cdcvg plugin? Either that or in my nginx redirect rules.
Look at the "protocol flow" here and you'll see actually to pay a zap, you need to make TWO get requests... https://github.com/nostr-protocol/nips/blob/master/57.md